Old Tape now uses your settings: Noise 10%, tone 75%, follow 50%, Hum 10%, Wow 50%, Flutter 5%, speed ×0.50, Drift 50%, Saturation, Age and Head bump 25%, Dropouts and Wear 20%.
Old Tape is first in the picker (Old Tape, Cassette, VHS, Clean), and newly added Tape effects start with it.
Old Tape now uses Clean's neutral base tone, so the preset sounds exactly like the settings you tuned on Clean.
Cassette, VHS and Clean presets are unchanged. Existing Tape effects keep their saved settings.
Verified 12 Tape tests in the simulator, including the exact Old Tape values. Not yet listened to on a physical iPhone.
Old Tape and Clean replace Reel and Micro. Choosing a preset now loads all of its slider settings; undo restores the previous ones.
Old Tape: slow, irregular pitch sag of a worn tape, with flutter, darker hiss, saturation, dropouts and uneven level.
Clean: every effect at zero with all sections on, so you start from untouched audio and raise only what you want.
Tape effects saved with Reel or Micro in builds 429–430 open as Cassette or Old Tape, keeping their slider values.
Verified 13 Tape tests in the simulator, including the Clean preset leaving audio untouched. Not yet listened to on a physical iPhone.
Build 430 · 1.0
Wow now reaches about ±65 cents and Flutter about ±150 cents at 100% on Cassette (previously 35 and 12). Micro and VHS go further.
Drift now makes the wobble irregular without weakening it. Wobble speed changes only how fast it moves, from slow drift to fast warble.
Fixed Flutter, which was not being applied. Defaults are now clearly audible (Wow 30%, Flutter 15%).
Depth and speed changes glide over a quarter second, so moving sliders doesn't swoop the pitch. Tape's fixed delay is now 24 ms, to make room for slow, deep wow.
Verified 12 Tape tests in the simulator, including measured pitch depth. Not yet listened to on a physical iPhone.
Build 429 · 1.0
Audio → + → Tape adds one effect with 19 controls: Type (Cassette, VHS, Reel, Micro), Amount, Mix and Output.
Noise section: hiss level, tone, ducking/following dynamics and 50 Hz hum. Wobble section: wow, flutter, speed, drift and stereo wobble.
Tape character (saturation, age, head bump) and Damage (dropouts, wear, mono) sections, each with its own on/off switch.
Works on tracks and Master, alongside existing effects, with undo, reset, reordering, saving and export. Adds a fixed 7 ms delay while enabled.
Verified 11 new Tape tests plus existing audio effect, merge, mix and undo regression tests in the simulator. Not yet listened to on a physical iPhone.
Build 428 · 1.0
AI Editing, audio-only editing, added cuts, final selection, AI Multicam, AI colors and narration scripts now request google/gemini-4-argon with high reasoning.
Uses OpenRouter’s reasoning.effort field and requires providers to support the requested parameters.
Copied AI request reports now show the reasoning effort alongside the requested model.
The specialized Gemini text-to-speech model continues generating narration audio.
Verified 34 simulator regression tests, including outgoing Argon requests, error reports, narration recovery, and single/multi-part editing with a synthetic video. No physical-device or paid live-generation tests performed.
Build 427 · 1.0
Added Old Armenian Movie to AI Audio’s narration styles. Selecting it automatically sets Armenian and pt-pt-storyteller-6; you can still change language or voice manually.
Uses the supplied Հայֆիլմ-style Armenian writing instructions and slow, deep, emotional delivery, with shorter scripts that leave room for pauses.
Sends the exact 1970s Soviet Armenian film-score description and the requested music exclusions through the provider’s negative_tags field.
Keeps Roast mode, optional direction and editable scripts. Saved soundtracks and resumed multi-part jobs retain the preset’s delivery and music settings.
Verified preset selection, voice/music request contents, recovery and existing narration workflows in the simulator. No physical-device testing performed.
Build 426 · 1.0
Deleting the Project Files category now removes every saved project and its referenced media, then opens a fresh empty project.
Includes shared project media, saved voiceover takes and files retained only for undo.
Keeps Unused Files, Exports, saved looks, settings, API keys and originals in Photos or Files.
Saves the project-library change before deleting media, preserving projects and files if the archive cannot be saved. The confirmation explains the full deletion scope.
Verified 42 simulator regression tests, including deletion, save-failure recovery and reopening projects. No physical-device testing performed.
Build 425 · 1.0
Fixed 23 audit findings across timeline editing, audio processing, AI recovery, project storage and rendering.
Kept continuous multicam microphone ownership, partial mutes and restoration handles consistent through track moves, merging, audio separation, speed changes and reverse.
Preserved narration-driven music ducking and prevented processed voiceover takes from doubling during review.
Allowed retries after definite AI-provider rejections and repaired multipart narration recovery after interrupted saves.
Moved routine project saves off the interface thread, preserved newer grades after Undo, fixed animated title rotation and cleared saved comparison looks consistently.
Preserved manual tracked-mask corrections, fractional frame rates, short proxy tail frames and transparent NightShot previews; mixed-format media now uses safe fallbacks.
Verified 342 distinct simulator regression tests; no physical-device testing performed for this release.
Build 424 · 1.0
Fix all 27 confirmed findings from the build 423 audit, with simulator regression coverage.
Recover completed AI responses and accepted jobs after local save failures, and keep rejected prompt changes from replacing saved jobs.
Preserve retained microphone audio through multicamera trims, cuts, duplication, mute and cleanup, including saved mix levels and detached narration.
Correct shared-bus compressor gain analysis and keep detached ducking tied to its intended narration source.
Keep tracked masks aligned with variable-frame-rate pictures and creative-effect renders; preserve alpha geometry, enabled media streams and fractional frame timing.
Load and duplicate large tracked projects without blocking the interface, keep photo-export cancellation responsive, and reuse unchanged prepared audio after timeline edits.
Save title-duration changes to the edited title and preserve Redo after unchanged AI Mix or track-name interactions.
Keep camera format choices consistent after lens fallback and preserve live-stream recovery through repeated slow encoding. Verified with simulators and local fixtures; no physical-device testing.
Build 423 · 1.0
Fix 27 confirmed audit issues with permanent regression coverage. Storage cleanup preserves newer saved edits, completed AI analysis survives relaunch, and deleted projects release orphaned recovery records.
Preserve detached narration, music and microphone routing, compression, filtering and ducking through repeated detach, track moves, audio separation, deletion and timing effects.
Improve multicam synchronization across multiple audio streams and silent dropouts. Restore trimmed microphone edges with the picture and retain speech timing metadata.
Correct tracked masks on reversed and variable-frame-rate footage, anamorphic and clean-aperture display geometry, transparent Retro Camera overlays and final-frame previews.
Make each audio-slider drag one Undo across track cards, retain Redo after unchanged edits, release unreachable title thumbnails, and clear recovered live-view connection warnings.
Protect interrupted narration and voiceover recovery, refresh changed music jobs, and keep playback muted after headphone removal. Verified using simulators, rendered media and browser recovery tests; no physical-device testing in this release.
Build 422 · 1.0
AI Video Edit now unlocks the prompt and selection after a failed, cancelled or expired provider job. Keep the draft when reopening, then change instructions and confirm a fresh attempt.
Interrupted requests and downloads resume the existing job. An explicit restart option allows revised instructions without silently creating duplicate paid jobs.
Include all 33 completed simulator-audit fixes covering project cleanup and export protection, large preset and storage responsiveness, timeline gaps and Undo, multicam sync and retained microphone timing.
Preserve detached audio compression, ducking scopes and effect order across preview, render and export. Keep mute editable after multicam rendering, and stop incompatible partial merges before changing the edit.
Fix render cancellation, transparent VHS overlays, neutral Retro Camera on Log footage, fractional final-frame seeking and quick fullscreen grading gestures. Correct native audio arithmetic and thin-frame VHS processing.
Guard late camera/live-sharing callbacks and malformed or refused AI responses. Verified with simulator regressions, native memory/arithmetic checks and an iOS build.
Build 421 · 1.0
Add Low & High Cut to the audio effect collection, with independently adjustable low and high cutoffs and bypass.
Add Ducking with a Driven by track selector, amount, attack, release, hold and lookahead. Audio-only tracks and video sound can drive it; muted or missing sources bypass safely, and circular routing is prevented.
Detach AI Audio now preserves narration low cut and original/music ducking as editable track effects. Keep high cut off, retain detected speech timing, and keep imported settings attached to their clips as they are moved, trimmed or split.
Use the same ducking envelope for preview, export, track rendering, merge and compressor gain analysis. Preserve existing echo/reverb tails when adjusting ducking.
Build 419 · 1.0
Opening Grade with an audio clip selected restores the last selected video in that project and moves the preview into that clip if needed. An already selected video stays selected.
Remember the video separately for each project, including after reopening or duplicating a project. Selection remains viewing activity and does not count as a content edit or add an undo step.
Keep the existing grading layout. If there is no usable remembered video, use the existing alert to ask you to select a video in Edit, instead of leaving inactive sliders on screen.
Prevent a queued preview reload after Undo from overriding a newer grading selection.
Build 418 · 1.0
Add Detach AI Audio to the track menu when the track contains AI narration or music. Create independent AI Narration and AI Music tracks while keeping the original footage and its sound in place.
Preserve clip timing, gaps, mix levels, mute settings, mono/stereo channels, looping music and fades. Detaching keeps extended narration footage and existing picture edits intact.
Remove AI ducking and narration low cut on detachment. Existing track audio effects remain editable on the separated tracks. Undo restores the combined AI audio and its settings.
Prepare and save the detached audio before changing the timeline. Cancellation, missing media or a failed save leaves the original project intact.
Build 417 · 1.0
Protect saved projects from stale editor sessions and keep large tracked-look autosaves off the interface thread. iPad uses a single editor window.
Keep audio that starts before or ends after picture, preserve source frame rate, and cancel export during audio preparation as well as rendering.
Fix timing in VHS, Retro Camera and Voice Cleanup previews. AI Color changes no longer remain hidden under conflicting keyframes, and animated titles move from their displayed position.
Preserve inherited fades across multicam cuts. Narration extensions remap fades, blends, mutes, keyframes and tracked masks, and removing an extension restores the original timing.
Preserve native transparency when converting AI source footage to 24 fps, extending narration footage, and using Retro Camera at zero Mix.
Keep saved voiceover audition independent of a moved timeline copy and stop pending playback safely when the audio route changes.
Limit simultaneous AI narration and music requests, retain completed sections on failure, and resume only missing work. Existing AI Audio controls remain accessible after short cuts or audio separation.
Clean up visual metadata when clips become audio-only and support valid short audio fragments. Normal imports now honor Startup Look and Randomize on Import.
Restore all camera-preview and live-view defaults when resetting settings, show edited titles in project thumbnails, and rebuild missing temporary photo media after storage cleanup.
Build 416 · 1.0
Adjusted deletion to require 1.5 times the original drag distance, reduced from build 415's doubled distance. The split threshold is unchanged.
The same distance applies to selected gaps. Releasing applies the highlighted action; dragging back up still cancels.
Build 415 · 1.0
Doubled the red delete threshold, from 1.5 to 3 clip heights down from the selected clip’s top. The yellow split threshold remains halfway down the clip.
The same distance applies to selected gaps. Releasing applies the highlighted action; dragging back up still cancels.
Build 414 · 1.0
Avoid rendering an entire recording’s audio just to open a neutral clip for grading. Equivalent source audio is reused across cuts, including when an active player’s cached audio exceeds the normal cache budget.
Use a lighter playback path for neutral audio while preserving clip volumes, gaps and cut fades. Audio effects and live meters remain active when needed, including when switching between Audio and Edit.
Passed 18 simulator checks and 8 optimized iPhone checks using the actual long recording. Grading loads measured about 0.2–0.26 seconds, down from 1.7–2.6 seconds; repeated selection and playback also improved.
Build 413 · 1.0
Preserve precise saved-file timing for Time & Pitch and Speed & Reverse. Imported audio and Voiceover recordings no longer fail the duration check because the MOV container rounded away a millisecond.
Duration checks remain in place. Genuine duration failures now explain the timing problem instead of suggesting a smaller image.
Reproduced the exact error with an imported MP3 before the fix. Passed 20 simulator regression checks covering pitch, fractional speeds, Voiceover audio, previews, export, Undo/Redo, cancellation, linked audio, multicam sound and Log video. No physical phone was used.
Build 412 · 1.0
Added Record Voiceover to Add Media and the top Add menu when an audio-only track is selected.
Record at the playhead and add the chosen take to that same track. Empty projects work too, with an audio recording screen instead of a blank video preview.
Keep the three-second countdown, multiple saved takes, take review, headphone playback, and optional wired microphone monitoring. Playback stays muted when headphones are absent.
Recording can continue beyond existing footage. It stops before the next clip on the destination track, and occupied positions cannot overwrite existing audio. Adding a take preserves track settings and supports undo/redo.
Passed 31 simulator checks, including actual microphone capture, retakes, gap boundaries, failed saves, take review, and the original voiceover workflow. No physical phone was used.
Build 411 · 1.0
Added Multicam alongside One after another and Separate tracks after selecting several media files from Photos, Files, or Seedgrade Storage.
Multicam uses local audio synchronization for two to four videos, preserves full recordings, reuses empty compatible tracks, and keeps existing footage in place. Cameras can align through another camera with overlapping sound.
Imports are saved as one undoable edit. Missing audio, uncertain matches, cancelled analysis, or failed saves leave the project unchanged and remove staged copies. Separate tracks remains available for manual alignment.
Improved synchronization to reject nearly constant audio whose tiny codec artifacts could produce a false match.
Passed 18 simulator checks covering two- and four-camera alignment, unrelated/silent audio, cancellation, saving, undo/redo, Photos transfer ownership, and existing import choices. No physical phone was used.
Build 410 · 1.0
Added all four tools to the More menu for audio-only clips in both timeline and clip-card views.
Time & Pitch and Voice Cleanup provide audio previews without requiring or generating video frames.
Audio results keep their track placement and media type, with editable track effects, saved projects, Undo and Redo. Voice Cleanup also preserves the clip mute switch.
Checked audio processing, previews, trim ranges, pause removal, export, cancellation and linked video/audio workflows in the simulator. No physical phone was used.
Build 409 · 1.0
Selecting or scrubbing between cuts keeps the prepared timeline player in place instead of reopening each cut as a separate grading source.
Grading and source-dependent clip tools prepare the selected source when needed, with cancellation protection when leaving the inspector.
Repeated cuts share cached source assets and loaded metadata. Replaced or deleted files invalidate the cache, and grades and audio mixes remain independent.
Validated in the simulator using an eight-minute source with 24 cuts, plus playback, scrubbing, grading, audio, gaps, project restore and export checks. The reported screen recording and a physical phone were not used.
Build 408 · 1.0
Audio → Record Voiceover records narration from the playhead while showing the timeline, with a silent three-second countdown and an accessible Stop button.
Playback is automatically muted during recording unless headphones are detected. Wired headphones can also monitor the microphone; Bluetooth headsets play the timeline without delayed self-monitoring.
Record multiple takes, review each against the timeline, and add the chosen take on a new audio track at its recorded position. Other takes remain saved with the project.
Headphone disconnection, audio interruption and playback stalls stop the take safely and preserve the recorded portion. Failed project saves offer Retry saving without rerecording.
Take storage survives project reopen, duplication and timeline undo. Removing a take from its list preserves copies already used in the timeline.
Validated in the simulator with microphone capture, timing and mixed-audio checks, route-loss simulations, save-failure recovery and compact layouts. Physical headphone routing and monitoring latency were not tested on a device.
Build 407 · 1.0
Mute clip now silences generated narration and music as well as source audio. Unmuting restores the existing mix settings.
Speed, reverse and Time & Pitch keep continuous multicam microphones synchronized with the picture and later clips. The Time & Pitch preview includes the microphone mix.
Render Clips, VHS and Retro Camera preserve the original edit if saving fails. Completed visual effects keep separately linked audio attached.
Changing the duration of a split title or still updates inherited fades to fit the new duration.
Trimming linked audio keeps the picture fade aligned with its source timing.
Deleting gaps or clips preserves short audio fragments outside the removed interval.
Verified with simulator regression tests, additional linked-effects checks and a Release build. No physical-device testing was performed.
Build 406 · 1.0
Render Clips and adjacent merges preserve transparency in imported images and video instead of filling transparent areas with black.
Track audio effects render continuously across cuts and gaps, retaining echo and reverb tails. Later renders, undo, reopening and cancellation preserve the edit.
Clip fades follow speed changes, reverse playback and trimming without disappearing early.
Splitting titles preserves their original fade timing instead of adding fades at every internal cut.
Ordinary grading, fullscreen adjustments and resizing retain existing keyframe easing when only a value is changed.
Verified with 122 simulator checks, additional transparency regression checks and a Release build. No physical-device testing was performed for this release.
Build 405 · 1.0
Local Masks: track a subject's position and size forward, backward or in both directions, using local analysis of the original video.
Review the result in the fixed preview, correct position, size or rotation, and track again from that moment. Loss of confidence stops tracking and marks the last confirmed position for review.
Tracking stays editable and persists with projects and presets. Trims, splits, speed changes and reversing preserve its timing; merging bakes it once.
Analysis can be cancelled without changing the existing mask and stops when the app leaves the foreground. Supports up to 30 minutes per analysis; automatic rotation and shape deformation are not included.
Build 404 · 1.0
Clip More menu: editable keyframes for placement, opacity and primary grading; local ellipse, rectangle and polygon masks with feathered adjustments.
Preview tools: waveform, RGB parade, vectorscope and editing proxies. Final exports continue using original media.
AI Multicam: synchronize two to four tracks from shared audio locally, review the offsets and apply with undo. Up to 30 minutes; ambiguous matches are rejected.
Animation follows trims, splits, speed changes and reversed footage. Masks and animated grades persist with projects and presets. Simulator validation covers rendering, export, cache cancellation and audio synchronization.
Build 403 · 1.0
Auto microphone can retain a working input when iOS cannot apply the preferred microphone. Explicit microphone selections remain verified.
Corrected Log preview pacing and removed an extra image copy. Preview rendering no longer waits for UI delivery.
Added Copy camera report in Camera settings and camera errors, with microphone routing and frame-drop diagnostics for troubleshooting.
Recording quality and stabilization settings are unchanged. Hardware confirmation of the reported 4K recording stutter is still needed.
Build 402 · 1.0
Preview stabilization now defaults to Off in Camera settings. Existing saved On or Off choices are preserved.
Recording stabilization is unchanged. Reopen Camera after changing the preview toggle.
With preview stabilization off, Both mode uses the native preview and Log appears flat. Single-camera Wi-Fi preview may use a lower resolution.
Build 401 · 1.0
Preview stabilization is on by default. Turn it off in Settings > Camera, then reopen Camera to compare preview delay while keeping recording stabilization unchanged.
Single-camera Log previews keep color conversion with preview stabilization off. This uses preview-sized buffers; Wi-Fi live view shares this output and may run below its requested resolution.
Both mode uses the native camera preview when preview stabilization is off, so Log appears flat. Shared recording buffers and recording stabilization remain unchanged.
Validated preference persistence, recording-connection isolation, camera mode switching, Log preview rendering, and live sharing in simulator tests. Physical camera latency awaits device testing.
Build 400 · 1.0
AI video edits are acknowledged only after the project is saved. Failed writes keep the original edit and generated result available for retry.
Unfinished AI Video Edit jobs protect their source, downloaded result, and converted-original files from unused-file cleanup, including jobs in other projects.
Ordinary high-frame-rate imports keep their original frames. Slow-motion normalization is restricted to actual retimed footage.
Ripple deletion removes the matching interval from continuous multicam audio, keeping picture and microphone sound aligned.
Linked picture and sound receive speed, reverse, and Time & Pitch timing changes together. Both renders must finish and save before either original is replaced.
Use AI audio now silences the corresponding interval of separated linked original sound, without muting unrelated tracks.
Remaining continuous sound becomes editable audio clips when its picture lane is empty. Overlapping microphones retain their simultaneous timing on separate audio tracks.
Waveforms refresh after ducking, interval mute, whole-mix mute, or gain changes. Existing drawings remain visible while audio-only updates load.
Build 399 · 1.0
Fixed damaged rows in full-resolution live camera frames by matching the monitor output’s stabilization to the recording output. Recording settings stay unchanged.
Live encoding now uses a bounded two-frame pipeline and accounts for hardware latency, avoiding unnecessary quality resets while the browser is waiting for its first picture.
Added hardware checks for matching stabilization and a regression test that verifies the stream becomes playable without repeated encoder restarts.
Build 398 · 1.0
Live View now waits for playable video before starting and reconnects automatically after camera, orientation or streaming-quality changes. Each format uses a fresh browser player to avoid carrying over parts of the previous image.
Corrected stream timing for 24 fps when the camera runs at 30 or 60 fps, and aligned the rendered image to the exact output dimensions.
Fixed a video-writer teardown crash found during hardware testing. Verified 4K picture edges in portrait and landscape, browser format changes, and local streaming with recording.
Build 397 · 1.0
Added Settings → Camera → Live view: Auto, 720p, 1080p or 4K, and Auto, 15, 24, 30 or 60 fps. Auto aims for 1080p/30.
Replaced the browser preview’s JPEG stream with H.264 video and a bundled local player. Streaming quality is capped by the actual camera frames and adapts to encoding load, temperature and connection performance. Recording settings are preserved.
The sharing screen shows the current stream quality. Video remains local and silent, with a short viewing delay. Orientation, selfie mirroring and browser full screen are retained.
Build 396 · 1.0
The browser preview switches between portrait and landscape even while the app interface stays in portrait, including during recording.
Keeps the selected aspect ratio and correctly orients the selfie camera and dual-camera inset. Recording orientation is unchanged.
Build 395 · 1.0
Moved Done to the top left of Live view.
Added a Share button at the top right for the active live-view link, using the standard iOS share sheet. The button is disabled until a link is available.
Build 394 · 1.0
Simplified the cut-point response format sent to Gemini after reproducing its HTTP 400 rejection on the connected iPhone. Timing, cut-count and full-clip preservation are still checked in the app.
Handles millisecond rounding at the clip endpoints without accepting genuinely out-of-range cuts. Copy report now includes the offending timestamp if a cut is invalid.
Verified with the original 182.48-second audio on the phone: the live request succeeded, produced 36 clips with full coverage, played correctly in an isolated test copy, and Undo restored the original. The user project was unchanged. All 20 simulator regression checks passed.
Build 393 · 1.0
Cuts only adds editable split points while keeping the entire clip and its original order. Target length and rearrangement are hidden; the clip count starts on Auto.
Send audio only analyzes sound without uploading video and adds an editable musical-beat suggestion without replacing existing instructions. AI Editing is also available on audio-only clips, with audio analysis always enabled.
Cuts preserve original media, grades, timing and Undo. Multipart analysis does not add cuts at arbitrary part boundaries. Verified with 19 simulator checks covering real audio files, silence, gaps, trimmed sources, UI controls, editing, playback and export. No physical phone tests or live paid AI requests were used.
Build 392 · 1.0
Projects with only regular footage show Standard (SDR). Adding Log footage reveals its working space; removing the last Log clip returns the display to Standard (SDR).
Your preferred Log working space is remembered separately from the displayed SDR state. Switching the recording profile to SDR no longer erases that preference or changes existing project grades.
Source metadata determines the displayed processing path. HDR and unreadable media are identified separately. Saved grades, footage appearance and recordings are unchanged. Verified with 23 simulator checks and real synthetic Log/SDR files; no physical phone was used.
Build 391 · 1.0
Replaced the segmented Tone Curve response with one smooth curve and broader, overlapping tonal adjustments. The graph, preview, video renders and LUT exports all use the same response.
Dragging a Tone Curve slider now displays only the curve line over the picture. The overlay follows the image dimensions, ignores touches and disappears when the adjustment ends or the panel closes. It is never included in exported footage.
Tone Curve still survives randomization and saves with projects and presets. Log highlight headroom is preserved. Verified with simulator rendering and interaction checks; no physical phone was used.
Build 390 · 1.0
Added Tone Curve after Film Gate with Shadows, Darks, Lights and Highlights sliders, a live curve graph, an on/off switch and reset controls. Double-tap a slider to reset it.
Tone Curve values and enabled state stay unchanged during randomization. The curve saves with presets and projects, participates in undo/redo and copied grades, and starts neutral in older projects.
The curve is applied after the randomized grade and its Mix, before the final Log display conversion. Extended Log highlights remain available for grading. Preview, video rendering and exported LUTs use the same curve. Tested in the simulator without using a physical phone.
Build 389 · 1.0
Camera settings now include Log preview conversion. It is on by default and displays Apple Log and Apple Log 2 with natural contrast and color while shooting. Turn it off to see the original flat preview.
This is a display-only conversion. Recorded Log files, recording quality and the selected editing color space stay unchanged. The preview follows camera rotation, mirroring, framing, full screen and dual-camera placement.
Preview processing is bounded to one frame set, at up to 30 fps and a 1280-pixel long edge. Unsupported preview outputs fall back with an explanation, keeping recording available. Verified with 51 simulator checks and synthetic Log frames; no physical phone was used.
Build 388 · 1.0
Selecting Apple Log or Apple Log 2 in Camera settings automatically selects ACEScct for editing. Standard SDR, Wide Color SDR and HDR select Original / No conversion; non-Log media keeps its existing display processing.
You can change the editing color space afterward. Your override stays in place when reopening settings or changing codec, resolution or FPS, until you explicitly select a recording color profile again. Restoring camera defaults also resets the editing choice to the standard recording default.
The selected conversion applies to the current project's video grades and the default for new grades. Unsupported selections and busy editing operations cannot partially change the recording/conversion choices. Tested in the simulator without using a physical phone.
Build 387 · 1.0
Camera settings > Editing color space now includes Original / No conversion. It preserves the encoded Apple Log or Log 2 appearance without applying the automatic gamut or display conversion, while keeping grading controls available.
The choice is saved with each grade and used for playback, original comparison, thumbnails, exports and color LUTs. Switching back to ACEScct, DaVinci Intermediate or ACEScg restores managed grading. ACEScct remains the default.
Original files remain unchanged. Exporting with No conversion preserves the flat appearance in the app's SDR output; it does not create a color-managed Rec.709 conversion or a tagged Log delivery file. Tested with synthetic media in the simulator, without using a physical phone.
Build 386 · 1.0
Camera settings now offers ACEScct (default), DaVinci Wide Gamut / Intermediate, and ACEScg Linear. The choice changes Log grading behavior while preserving scene highlights before the SDR display conversion.
Changing the setting applies it to this project's video grades and remembers a default for new grades. Saved projects and presets retain their choice, older projects retain ACEScct, and grading reset preserves the selected working space. Preview, exports and color LUTs use the saved choice.
Recording profiles and original media remain unchanged. Existing SDR grading remains unchanged; rendered output is still SDR. Verified using synthetic Log footage and simulator checks, without using a physical phone.
Build 385 · 1.0
Recognized Apple Log and Apple Log 2 footage automatically uses an ACEScct working space, with the correct gamut for each source. Gain and white balance adjust scene values before the final SDR display conversion, retaining recoverable highlights in the original footage.
Paused grading, playback, thumbnails, layered edits, rendered exports and exported color LUTs use consistent Log interpretation. Original recordings remain unchanged; SDR projects keep their existing grading behavior. Merged, flattened and exported video is still SDR.
Verified with locally generated, tagged 10-bit Log footage in the simulator, including highlight recovery, preview/export agreement, SDR metadata, alpha and existing grading/export regressions. No physical phone was used.
Build 384 · 1.0
Replaced tmpfiles.org with Uguu: no signup required, direct HTTPS video links and automatic expiry after three hours. Updated the upload confirmation accordingly.
Temporary connection failures retry once before AI submission. Copy report now identifies the upload host, attempt count, HTTP status and underlying network error without including private media links.
Eight upload and saved-job regression checks passed, including a real iOS-networking upload and byte-for-byte video download verification. Paid OpenRouter generation was not tested because the available simulator has no configured API key; Ed iPhone is unavailable.
Build 383 · 1.0
Upgraded narration generation to google/gemini-3.8-flash-tts. Spoken text is sent in input; tone, pacing, performance and cleanly formatted timing guidance are sent through speech_metadata.style.
Keeps the existing 30 voices and creative-direction controls. Multi-part narration shares its delivery style, and optional extra narration sends only its own words. Existing generated audio is retained.
Verified main and continuation generation through OpenRouter on Ed iPhone, including playable 24 kHz mono PCM audio. Request, custom-text, multi-part and saved-job resume regression checks passed.
Build 382 · 1.0
The local live-view page has a small fullscreen toggle at the bottom right. It enters and exits fullscreen where supported, and explains when a browser restricts fullscreen.
The camera Wi-Fi icon is white while sharing is off. Its own dot is green while sharing is ready; when viewers connect, green cycles through the dot, inner arc, and outer arc. The separate badge is removed.
Reduce Motion keeps a steady green dot. Sharing and camera lifecycle checks, browser fullscreen checks, and device recording tests are included in validation.
Build 381 · 1.0
A Wi-Fi button at the top-right of the camera preview opens Live view sharing. A small green dot appears beside the icon while sharing is active.
Scan the QR code or copy the private local link to a browser on the same Wi-Fi or Personal Hotspot. The viewer shows only the feed; Both mode includes the selfie inset. No cloud upload or audio streaming is used.
The monitoring stream targets up to 720p and 12 fps, supports four viewers, and drops frames for slow connections rather than queueing recording work. Recording retains its selected settings. Start sharing before a single-camera take.
Sharing stops when leaving Camera or backgrounding the app. Each new sharing session receives a fresh link. Tested on Ed iPhone while recording and switching cameras, and viewed from the Mac over Wi-Fi.
Build 380 · 1.0
Merged video tracks and AI-generated clips now keep neutral grading controls enabled, so general color randomization responds immediately. The image remains unchanged until you grade it.
Existing saved merged and AI-generated clips with the affected neutral state recover their grading controls when reopened. Dormant nonneutral grades remain disabled.
Verified AI overlay, mixed-frame-rate track merging, live grading, export composition, saved-project reopening, and existing merge cancellation/undo behavior.
Build 379 · 1.0
The first imported or recorded video establishes the project’s default export resolution and frame rate. These remain stable after trimming, rearranging or deleting that original clip.
While AI Video Edit footage remains in the project, export defaults to 24 fps, even if that generated footage was matched to another frame rate. Removing all AI footage restores the saved original rate.
Defaults are applied when opening export settings; manual choices remain editable for that export. Fractional rates such as 29.97 fps and nonstandard original dimensions are preserved, with available rates verified through the encoder.
Saved AI provenance survives renaming, splitting and merging. Existing projects without source metadata initialize from their first remaining original video in stored clip order.
Build 378 · 1.0
Convert original to 24 fps is on by default. Match original FPS is off by default; enabling either turns the other off. Both may be disabled.
Original conversion preserves the full source clock, trim points, playback speed, audio timing and editable effects. The original file stays untouched, and both timeline changes undo together.
Generated video can match the original frame rate, including 29.97 fps, with explicit frame repetition or removal. Saved AI jobs resume without another paid submission.
Build 377 · 1.0
Merging tracks after trimming and selecting a gap now clears the obsolete gap selection instead of reporting a false rendering failure.
Merged media is retained as soon as the new edit is committed. Preview changes or late cancellation cannot delete a file already referenced by the project.
Verified short AI-generated overlays over trimmed originals, reopening, undo/redo, audio mixing, and cancellation. Failed merges before commitment keep the original tracks intact.
Build 376 · 1.0
Clip → More → Fade In & Out adds visual fades with one shared duration slider or separate fade-in and fade-out controls. Zero disables an end.
Preview fades over the surrounding tracks before applying. Settings remain editable, save with projects, and support undo; audio remains unchanged.
Timeline playback and exports use the same transparency. Render Clips and merges bake fades once while preserving transparency for underlying tracks.
Build 375 · 1.0
AI Video Edit in clip More sends a selected 1–15 second interval and text instructions to FLUX Video Edit through OpenRouter. Includes preview, selection timeline, position/length controls and independent scrubbing.
Edited video is resized to the source dimensions and placed on a new track immediately above the original, without moving surrounding clips. Keep original audio by default, or mute only the matching original interval and use AI audio.
A confirmation shows estimated model cost and temporary tmpfiles.org hosting with one-hour expiry. Saved job IDs let interrupted downloads and waiting resume without submitting another generation.
Build 374 · 1.0
Device storage usage now accounts for cache space iOS can reclaim, with total, used and available amounts shown separately from Seedgrade’s media and files.
The storage gauge uses shorter ticks along the rounded-square button edges, retaining one usage color across lit ticks.
Add Media → Seedgrade Storage previews include a centered Select button. Selecting closes the preview and marks the file for import without duplicating or deselecting an existing choice.
Build 373 · 1.0
Short ticks around the storage icon show the entire device’s used storage percentage. Unfilled ticks remain dim.
All lit ticks share one color, progressing from sky blue to green, amber and red as storage fills.
Usage is read in the background and refreshed while Projects is visible, when returning to the app and after closing Storage. The button still opens Storage.
Build 372 · 1.0
Auto Style lets Gemini choose an original narration approach from the footage, beyond the preset styles, and carry it coherently through longer videos.
God of War adds original epic, solemn, mythic storytelling with deliberate pacing and grounded intensity, without copying game dialogue or imitating a particular actor.
Both styles support Roast mode, Your direction, the selected voice and language, and the existing narration time budget.
Build 371 · 1.0
The main camera zoom row uses the available screen width, including black margins beside a cropped preview, so native zoom buttons remain fully visible.
The fullscreen button keeps its own space, and zoom buttons retain their existing size and touch targets.
Build 370 · 1.0
Zoom controls and the fullscreen button use the current cropped preview bounds immediately after switching to Cinema, without needing a fullscreen round trip.
The shutter keeps its physical orientation while idle, so its recording animation starts in the correct direction when the phone is already landscape.
Verified Cinema control positions, fullscreen transitions, and Main, Selfie and Both recordings on iPhone.
Build 369 · 1.0
Cinema uses a tall 1:2.39 frame in portrait and a wide 2.39:1 frame in landscape instead of always cropping to landscape.
Applies to the live Main and Selfie previews, the dual-camera inset, and saved recordings. Each recording keeps its starting orientation.
Square and Portrait 4:5 keep their existing behavior.
Build 368 · 1.0
Adds Square 1:1, Portrait 4:5 and Cinema 2.39:1 beside Standard and Open Gate for Main, Selfie and Both camera modes.
Live previews match the centered crop saved to each recording. Portrait stays vertical and Cinema stays wide; pinch zoom, focus, exposure, inset positioning and fullscreen remain available.
Cropped takes finish in the background before timeline import, preserving microphone timing, the selected codec and color profile, and 10-bit detail. Original takes remain recoverable if finishing fails.
Build 367 · 1.0
Uses Core Motion gravity readings to rotate the inner recording stop button and timer independently of screen-orientation lock.
Keeps the control stable near diagonal angles and when the phone lies flat. Motion tracking stops when the camera preview closes or is suspended.
The outer shutter circle and the orientation of an ongoing recording remain unchanged.
Build 366 · 1.0
Reduced the enter-full-screen arrows from 18 to 14 points for a more compact appearance beside the zoom controls.
Kept the exit icon, circular background and touch target unchanged.
Build 365 · 1.0
The full-screen button now has a 30-point visible circle, matching the zoom capsules and keeping their centers aligned.
Kept a 44-point tap target and matched the exit-full-screen circle to the same size.
Build 364 · 1.0
Added a full-screen button at the right edge of the camera zoom row. Full screen shows the live feed, record/stop button and exit control, including both previews in dual-camera mode.
Entering and leaving full screen keeps the existing capture session and can be done during a recording without restarting the take.
The inner stop rectangle and recording timer rotate with the device while the outer shutter circle stays fixed. Recorded orientation remains locked for the current take.
Build 363 · 1.0
Added stabilization above aspect ratio, using OFF, AUTO, STD, CINE, EXT, EXT+ and FAST labels, with descriptive menu names and unavailable modes disabled.
Added stabilization to Camera settings. Standard framing defaults to Auto and Open Gate to Off; each framing and camera mode remembers its choice. Controls lock during recording.
Checks active recording connections, format support and dual-camera capacity, verifies the applied modes, and falls back to Off when a choice cannot run.
Preserves microphone audio arriving ahead of stabilized video and drains captured frames after Stop so dual-camera recordings keep their beginning and end.
Build 361 · 1.0
Moved the aspect ratio selector to the left of the frame-rate/resolution controls and the microphone selector to the right of Record, beside the camera switch.
Recording, frame-rate/resolution controls, microphone metering and selector behavior are unchanged.
Build 360 · 1.0
After a camera recording is successfully saved to the timeline, the Edit tab icon briefly flashes green and gives a small wiggle, then returns to its normal appearance.
Only the icon animates: the Edit label and tab background stay unchanged, and recording stays on the Camera page. The cue follows completed saves, including a successful retry, rather than the Stop button.
Reduce Motion keeps the green confirmation while suppressing the wiggle.
Verified successful saves, failed-save retry behavior and the hosted Camera workflow; checked the icon during its green cue and after it returns to normal.
Build 359 · 1.0
Tapping another tab leaves Camera immediately while capture shuts down asynchronously. Recording and saving still protect navigation until the take is safe.
Playback requests wait for camera audio resources to be released instead of failing during shutdown. Returning quickly to Camera queues startup after the previous stop, and old shutdown work cannot reset the new session.
Cancelled or superseded page-recovery work cannot reopen the Edit preview after navigating elsewhere. Inactive camera sessions ignore late interruption notifications.
Validated with 45 simulator checks and three real dual-camera recording/import/playback cycles on iPhone, requesting playback immediately after changing tabs.
Build 358 · 1.0
Camera entry now offers New Project and Keep Rolling with equal-width, equal-height buttons.
Explanations near the bottom clarify that New Project saves this project and starts recording in a new one, while Keep Rolling continues recording alongside existing footage.
The choice screen can scroll when space or larger text requires it. Existing camera-entry and project-saving behavior is unchanged.
Build 357 · 1.0
Replaced the Camera-entry popover with a centered camera icon and New Project / Camera buttons. Capture stays off until a choice is made; New Project saves the current project before creating an empty one, while Camera continues recording into the current project.
Returning to Camera after reviewing, playing, scrubbing or selecting footage opens it directly. Project edits bring the choice back, including changes to clips, tracks, grading, titles and audio, even when subsequently undone. Opening a different existing project also requires a fresh choice.
Completed recordings are included in the review reference, so recording and reviewing more takes does not repeatedly ask. Empty projects continue to open Camera directly. Project-creation failures and foreground recovery cannot bypass the choice.
Validated with 37 simulator tests, the hosted choice/review workflow on iPhone, and three real dual-camera recording/import/playback cycles that append to the same project without asking again.
Build 356 · 1.0
Camera frame-rate menus exclude all values below 12 fps, including native 2-fps endpoints. Native support checks and fallback selection use the same minimum, while retaining compatible higher-rate formats.
When entering Camera from an existing project, a neutral placeholder is shown and capture stays off until New Project successfully finishes. Dismissing the prompt returns to the previous tab. Empty projects open Camera directly; repeated recordings within Camera do not ask again.
Project-creation failure, duplicate confirmations, foreground recovery and late completion after leaving Camera cannot bypass the entry gate. Foreground state still updates on other tabs so later camera entry works normally. The camera FPS floor does not change export rate selection.
Validated with 43 simulator checks and three device tests: project-entry confirmation/cancellation, three dual-camera recording/import/playback cycles, and every advertised Selfie 1080p frame rate recorded in both HEVC and H.264.
Publishing now reuses installer design validation when the design is unchanged, retries only stale website assets, allows time for pCloud public-link propagation and retains signed artifacts for hosting retries.
Build 354 · 1.0
Camera FPS menus now include compatible alternative native formats at the selected resolution, rather than only the format currently running. Choices remain filtered by camera mode, framing, codec, color profile and bit depth.
Dual-camera FPS menus use compatible front/rear format pairs within the selected lens group. Selecting a rate reconfigures the real sensor format and retains safe startup checks and fallback; no device models or resolution overrides are hard-coded.
Includes the confirmed HDR-setting crash fix from the directly installed build 353: legacy HDR setters are only called when the active format supports them.
Validated on the connected iPhone 16 Pro with real saved recordings: Main up to 4K120 and 1080p240, Selfie up to 4K60 and 1080p120, and the tested dual-camera group up to paired 4K30 and 1080p60. Verified every advertised Selfie 1080p rate with both HEVC and H.264, and repeated transitions back to Main. These are measured capabilities on this device, not fixed limits for other devices.
All 63 selected simulator regression tests and both physical recording test suites passed. Disposable recordings were deleted and saved camera preferences restored.
Build 352 · 1.0
Single-camera shutdown now detaches inputs, outputs and connections before another session claims the cameras. Returning to Main creates a fresh movie output, and shutdown waits for an active recording to finish before detaching its graph.
Dual-camera graph configuration, startup, cleanup and lens-group control restoration now contain native configuration exceptions and return recoverable errors. Failed cleanup prevents a new camera graph from claiming the same devices.
Dual-camera video dimensions are assigned after connection rotation, and previous output settings are cleared. Startup uses bounded compatible-setting fallbacks instead of probing every FPS configuration; displayed FPS choices come from the active format pair.
84 simulator regression tests passed; two hardware-only checks were skipped. The iOS build succeeded. A connected iPad test launched but skipped because camera/microphone permission was unavailable. The affected iPhone remains unavailable, so the reported crash has not been reproduced or verified fixed on that hardware.
Build 351 · 1.0
Preview layers, rotation observers and zoom observers now disconnect before the capture session is rebuilt, then reconnect after startup completes.
Removed startup probing that repeatedly configured every possible frame rate. The FPS menu now uses the active sensor format that actually started.
Single-camera startup follows a bounded fallback plan for frame rate, resolution, codec and color settings, including a standard SDR 8-bit option. Actual successful settings are saved and reflected in the camera controls.
60 simulator regression tests passed; one native-format hardware test was skipped. The iOS device build succeeded without compiler warnings. The affected device was unavailable, so its reported crash and live camera recording could not be verified on hardware.
Build 350 · 1.0
Single-camera FPS choices now require the complete recording graph to accept the device format, frame timing, color profile and movie encoder. Rejected combinations are omitted instead of being inferred from sensor ranges alone.
Format changes, movie encoder settings, session commit and startup now run through a native exception boundary. A rejected configuration is cleaned up and a compatible rate is retried instead of allowing an Objective-C exception to crash through Swift.
Frame timing and format are checked again after configuration and startup. Main and Selfie retain separate rate choices, and Camera settings discovery remains read-only.
59 simulator regression tests passed; five hardware-only checks were skipped. The physical-device test build compiled successfully. A paired iPad UI attempt was interrupted when the app was no longer foreground; the reported iPhone remained unavailable, so the exact reported crash and live recording remain unverified on that device.
Build 349 · 1.0
Camera frame rates now account for active camera-feature limits, retain native timing precision, and validate frame durations immediately before applying them. Unsupported native timing exceptions become recoverable errors instead of crashing the app.
Main, Selfie and Both remember their own negotiated FPS. Switching cameras selects a supported rate, and rejected settings safely retry compatible configurations. The 1 fps option remains excluded.
Video export no longer has a fixed 60 fps ceiling. Options extend to the source footage's frame rate and are checked with a short real encode/decode at the selected resolution, codec, container and bitrate. Unavailable combinations are omitted.
Export uses a precise video timebase for higher frame rates, and automatic bitrate scales with FPS. Capability checks are cancellable and successful results are cached.
52 simulator checks and 10 physical iPad checks passed. Real H.264 and HEVC exports at 120 and 240 fps were decoded and their frame timestamps verified on both. Live capture testing was skipped: the iPhone was unavailable and the paired iPad lacked camera/microphone permission.
Build 348 · 1.0
Automatic lens switching is a saved, global Camera setting, enabled by default. Changes apply the next time you open the camera. Resetting application settings restores it to on.
On supported virtual cameras, turning it off disables fallback lens changes for lighting and focus while preserving lens handoffs required by deliberate zoom changes. Physical single-lens cameras are safely left unchanged.
The preference applies to preview and recording in Main, Selfie and Both, including rear lens-group changes. Single-camera movie recording no longer overrides the device policy back to automatic.
45 simulator checks and seven physical iPad checks passed, including interacting with the toggle and reopening Camera settings. Capture testing was skipped because the paired iPad lacks camera/microphone permission; the multi-lens iPhone is unavailable.
Build 346 · 1.0
Opening Camera settings now reads advertised format metadata without creating capture sessions, locking devices or changing active camera format, color or frame rate. The previous hardware scan and restoration code have been removed.
Single-camera codec choices follow the source layouts handled by the recording path: HEVC/H.264 for supported 4:2:0 sources and ProRes for 10-bit 4:2:2. Actual movie-output codec availability is still checked for the chosen setup at camera startup.
Automatic setting compatibility, disabled incompatible choices and ProRes storage-time estimates are retained. Camera settings no longer require camera permission just to list capabilities.
45 simulator tests passed, with one device-only check skipped. Four physical iPad checks passed, including three openings of the actual Camera settings screen with saved Log/ProRes preferences and repeated native capability enumeration without camera state or permission changes.
The affected iPhone remains unavailable; the iPad checks do not establish iPhone-specific verification.
Build 345 · 1.0
Corrected HDR state restoration after camera capability discovery. Automatic HDR is now restored only after manual HDR writes finish, avoiding an AVFoundation exception.
The same safe restoration runs when a capability scan is cancelled. Codec compatibility choices and ProRes remaining-time estimates are preserved.
Forty-seven targeted simulator tests passed, including regression tests for the documented HDR setter preconditions, automatic/manual HDR restoration, cancellation, camera lifecycle, settings compatibility and storage estimates.
The affected iPhone was unavailable; physical camera verification was not performed for this fix.
Build 344 · 1.0
Color-profile selection now resolves codec, bit depth, resolution and frame rate as one supported recording combination. Single-camera capability discovery checks the configured movie output rather than assuming that every encoder supports every profile.
Unavailable manual codec and bit-depth choices are disabled. Compatible manual codec selections are preserved, including HEVC Log wherever the recording output advertises it. Previously saved incompatible combinations are reconciled when opening the camera.
ProRes storage time is estimated before recording using its data rate, actual resolution and FPS, with audio/container allowance and storage reserve. During a take, a higher measured file rate makes the estimate more conservative.
Forty-four targeted simulator tests passed. Physical recording checks were skipped because the paired iPad lacked camera/microphone permission and the affected iPhone was unavailable.
A broader existing gradient test reproduced identical color-tolerance failures on unchanged build-343 camera code; this is recorded separately from the passing targeted tests.
Build 343 · 1.0
Configure separate recording defaults for Main, Selfie and Both cameras in Settings → Camera → Recording settings.
Choose supported SDR, wide-color SDR, HDR HLG or Apple Log profiles, HEVC/H.264 or compatible single-camera ProRes, and available bit depths. Options are filtered by device and camera mode.
Set automatic or custom target bitrate and resolution/FPS defaults. Recording settings remain locked during recording and saving.
HDR and Log recording are available where supported; the editor and rendered exports still use SDR. Original recordings retain their capture profile.
Thirty-five simulator checks passed, including actual encoded HEVC/H.264 SDR and HEVC HLG files, metadata verification, independent settings persistence, compatibility filtering and camera lifecycle regressions.
Build 342 · 1.0
Storage and battery icons keep fixed positions with their left edges aligned.
Tap a timeline gap to select it. Drag the playhead triangle to yellow to split it, or red to remove it and close that amount of time on the selected track. Split portions stay separately selectable and support save, undo and redo.
Gap removal preserves other tracks, updates later gap boundaries and ripples continuous audio on the edited lane. Existing clip gestures and move mode retain their behavior.
Nineteen simulator tests passed, including native gesture/layout checks, saved gap splits, track isolation, real-media playback and export, and existing clip split/delete regressions.
Build 341 · 1.0
Storage and battery icons stay vertically aligned in a compact fixed column near the right edge, without moving as the time or status text changes.
Battery estimates remember the last three reliable takes across app launches, show an initial estimate from that history, and switch to the current take once measurements arrive. Long takes update one history entry; short takes and charging do not overwrite useful history.
The first one-point battery drop can now show an early estimate after one minute of recording, including 100% to 99%. Stronger two-point/two-minute measurements are retained for future takes.
The camera header shows approximate recording time available from storage, based on the active formats, frame rate and both files in dual-camera mode.
A second header row estimates remaining recording time from battery drain observed during the current take. It waits for sufficient readings and shows Charging while connected to power.
Title controls now scroll in a panel below the live preview instead of expanding over it. Adjust and position titles while keeping the preview visible, then tap Done.
Verified Main, Selfie and dual-camera recording on iPad mini; checked compact/tablet layouts, title editing, storage calculations and battery estimation logic.
Build 338 · 1.0
Pause unfinished timeline preview generation during playback and resume it after pausing. Existing thumbnails and waveforms remain visible.
Wait asynchronously for in-progress thumbnail decoding to stop before starting playback, with safe cancellation and retry when views reopen. Video and export quality are unchanged.
Verified repeated playback and thumbnail resumption on iPad mini with the reported high-bitrate 4K source; eight device tests passed.
Build 337 · 1.0
Detect truncated MP4/MOV media before effect rendering or export and explain that a complete original is needed, instead of failing later with Invalid sample cursor.
Render and export failures offer Copy error details with the failing stage, underlying error codes, export settings and clip ranges. Invalid trim ranges and export settings now have specific messages.
Verified the supplied portrait source through grading, rendering and export on iPhone. Its reported failure was not reproduced; diagnostic reports will help identify it if it recurs.
Build 336 · 1.0
Cancelling a timeline export now signals its background worker instead of synchronously stopping AVFoundation readers and writers on the UI thread. The worker tears down its own resources and removes the incomplete movie.
Single-clip video exports use a cooperative cancellation signal in their frame processor, covering cancellation during startup and preventing cancelled results from opening the share sheet.
The camera FPS selector omits 1 fps even when the device advertises it. Other supported frame rates remain available.
Build 335 · 1.0
The camera resolution caption now aligns with the microphone, framing and camera captions. A compact FPS selector sits directly above it within the same control height.
Frame-rate choices use supported standard cadences and native range endpoints exposed by camera formats at the chosen resolution and framing. Higher rates can select another compatible camera format; zoom choices refresh with that format.
Both mode validates common frame rates against the complete capture graph and hardware limits. Changing resolution, framing or lens group refreshes the FPS options and actual selected rate.
Sensor timing, encoding bitrate, dual-camera frame pairing and movie end timing follow the selected FPS. Both FPS and resolution remain locked throughout recording and saving.
Build 334 · 1.0
Choose recording resolution beside the microphone. Options are discovered from the current camera and framing formats at 30 fps, including 4K where supported. The default remains 1080p when available.
Both-camera mode offers common native resolutions validated against the complete capture graph and hardware resource limits. Recorded files retain the selected resolution instead of being capped at HD.
Open Gate preserves its sensor shape and labels resolution by native pixel width. Camera, framing and lens-group changes refresh the available options.
Elapsed recording time appears inside the record button only during a take. Resolution changes are disabled during recording, finishing and saving.
Build 333 · 1.0
Choose Titles from the new-track plus menu to add a separate transparent title track. New titles last five seconds, with adjustable timing, splitting, duplication and movement.
Drag titles directly in the preview, pinch to resize, rotate with two fingers and double-tap to edit. Center and edge guides help with alignment.
Customize typography, text color, alignment, outline, shadow, background, opacity and rotation. Plain title, Subtitle and Lower third styles remain editable.
Optional fade-in and fade-out durations render consistently in playback and export. Titles are saved with projects, support undo, and do not create audio tracks.
Build 332 · 1.0
Custom text is now the first AI audio style. Paste your narration separately from Your direction; Gemini adapts the text to the video while keeping its meaning. Roast mode remains available.
A local word-count guide estimates suitable text length from the footage duration. Text over ten times the guide shows an advisory warning without blocking generation.
Long-video requests include the full source, section timing and previous narration to guide a continuous adaptation. Saved jobs retain their original source text when resumed.
New narration prompts and speech timing use concise seconds with at most two decimal places. Media timing retains full precision.
Build 331 · 1.0
Drag downward on a storage video or audio preview to close it, including over visible playback controls. Preview playback stops and releases its media when closed.
Double-tap a grading slider to restore that control’s own default, with haptic feedback and undo. Other grading controls keep their values.
Camera recording waits through a short route-aware quiet interval after the start cue. Multi-camera capture discards queued samples from before recording is armed to keep the cue tail out of the take.
Build 330 · 1.0
Add Media now includes Seedgrade Storage. Select compatible stored media, preview thumbnails, and choose sequential or aligned separate tracks when importing multiple files. Originals remain in storage.
Swipe downward from the upper portion of a storage preview to close it and return to the file list.
Automatic gain compensation now runs only for overall grade randomization. Randomizing an individual effect module leaves unrelated grading controls alone.
Full-screen gesture adjustments show the current value beside Gain, Saturation, Contrast, Temperature, or Tint.
Build 329 · 1.0
Manual vertical scrolling can bring the last track to the top of the timeline. Selecting a clip still scrolls only enough to reveal it.
The vertical scroll range stays stable when the preview area changes, so a revealed lower track is not pushed back under the bottom panel.
Splitting keeps the timeline preview dimensions while playback rebuilds, even when the selected source has a different aspect ratio. The playhead triangle stays aligned with its selected track.
The camera start-recording cue now uses the supplied short tiny beep.
Build 328 · 1.0
Selecting a fully visible clip leaves vertical track scrolling unchanged. Hidden clips are revealed only as far as needed, with a small padding below the row.
The playhead triangle sits above the selected clip on its own track and follows vertical timeline scrolling.
Dragging past half the clip height previews Split in yellow; dragging half a clip-height below its bottom previews Delete in red. Releasing applies the action only to the selected segment; deletion closes the gap.
Dragging upward cancels the pending action and stops the triangle at its original position. Interrupted gestures cancel without changing the edit.
Build 327 · 1.0
Tap any other clip in Move mode to commit the current move and continue moving the new selection, including clips on other tracks. Each completed move has its own Undo step; switching without moving adds none.
A shared clip edge flashes green for one second when a move meets its left or right neighbor. It does not flash for the start of the timeline or unrelated tracks, and does not repeatedly flash while pushing against the same neighbor.
Dragging the selected clip moves it; dragging other clips or empty space continues to scrub. Playback, zoom, track positioning and the Done checkmark remain available.
Build 326 · 1.0
Selecting a clip brings its track to the top of the timeline viewport, including the last track, without changing track order or playhead time.
In Move mode, horizontal drags move the selected clip only when the drag starts on that clip. Other clips, track gaps, the ruler and lower empty space scrub the timeline normally.
Scrubbing previews the latest move draft and keeps the moving clip selected. Playback, pinch zoom, the Done reminder and one-step undo remain available.
Fixed a waveform preview audio-mixer crash reproduced when splitting a clip and undoing; waveform decoding now uses full-rate floating-point samples before reducing them to display peaks.
Build 325 · 1.0
Storage is now an icon immediately to the left of Settings on the Projects page. Its old Settings entry was removed.
Media files in Storage show square thumbnails without play overlays. Tap a thumbnail to open the video, photo or audio preview; tap the filename area for details and deletion controls.
Thumbnails load asynchronously with a bounded memory cache. Preview navigation includes a Done button, including when opened from search.
Build 324 · 1.0
Playback now previews the current clip position during Move mode. Further moves pause playback; Done commits one undoable change.
Timeline pinch-to-zoom works during Move mode, with subsequent drags using the new scale.
The checkmark gives a brief wiggle after a move gesture, respecting Reduce Motion.
The timeline quick menu is ordered Move, Split, Delete, More.
Targeted simulator tests and two native gesture tests passed for playback, pinch, timing, save isolation, undo and navigation.
Build 323 · 1.0
The timeline clip menu now includes Move alongside Split, Delete and More.
In Move mode, drag left or right on the timeline or the empty area below it to adjust the selected clip timing while the playhead stays fixed. The full-screen button becomes a checkmark to finish.
The clip stays on its track. Neighboring clips bound the move, linked audio/video stays synchronized, and Done commits one undoable edit. Leaving the timeline cancels an unfinished move.
19 targeted simulator checks passed, followed by 6 final regression checks and a native gesture test covering repeated moves, empty-area dragging, Done, Undo and tab navigation.
Build 322 · 1.0
Video requests now send literal slashes in MIME types and base64 data, avoiding the escaped video\/mp4 spelling rejected in a reported Google AI Studio response.
The shared encoding change covers AI Editing, Multi-Cam and other AI media requests without changing media bytes, prompts, models or retry policies.
12 targeted simulator tests passed, including inspection of the actual outgoing request, media-byte preservation, nested MIME error diagnostics, cancellation and failed-part handling. Live provider acceptance remains to be confirmed by retesting.
Build 321 · 1.0
AI Editing no longer shows unrelated music-submission advice for provider failures.
Copy report includes provider error details, HTTP status, request IDs, model, failed stage, part timing, upload size and edit settings when available. Credentials, media data and request prompts are excluded from the request summary; provider details are filtered and bounded.
Errors returned inside successful HTTP responses are detected. AI Editing and Multi-Cam requests retain their no-automatic-retry behavior. Cancellation stops immediately.
10 targeted simulator tests passed, including failures in the second analysis part and final selection. Tests used local response fixtures, not paid API requests. The original provider rejection still needs a new report to diagnose.
Build 320 · 1.0
Camera recording uses the supplied start and stop sounds in Main, Selfie and Both modes. Turn both off in Settings > Camera > Camera sounds; they are on by default.
The start cue finishes before capture begins. The stop cue plays once when capture finishes. The sounds use the current audio output without changing the selected microphone.
Rapid Start/Stop taps, interruptions and leaving the camera during the start cue cancel safely. Reset Settings restores Camera sounds to on.
38 targeted simulator checks passed, including cue playback, camera lifecycle, framing and microphone regressions. Ed iPhone was unavailable for a real-device audio check.
Build 319 · 1.0
After selecting multiple items in Photos or Files, choose One after another or Separate tracks before importing. Single-item imports work as before.
Separate tracks places each item on its own track at 00:00, reusing compatible empty tracks. One after another appends items in order to the selected track.
Cancel leaves the project unchanged. Failed imports do not create extra empty tracks, and the batch can be undone together.
Ten targeted simulator checks passed, covering video, photos, audio, track alignment, existing clips, saved projects, undo and native dialog dismissal. Physical-device import was not verified because Ed iPhone was unavailable.
Build 318 · 1.0
The camera microphone icon fills from bottom to top as input levels rise: green for normal levels, amber near the limit, and a brief red clipping warning. It works before and during recording.
The meter follows the captured input, including external microphones. In Paired mode it shows the louder microphone or channel, so a peak on either is visible. Recording audio is not adjusted by the meter.
The indicator rises quickly and falls smoothly, clears stale peaks after source changes, and stays bright while source selection is disabled during recording.
Automated PCM, meter timing, camera lifecycle and paired recording checks passed, including visual checks at a small phone width. Ed iPhone remained unavailable; live hardware metering was not verified.
Build 317 · 1.0
The microphone menu now includes external inputs reported by iOS, including devices with no additional data-source choices. Bluetooth microphone input support is enabled.
Auto prefers a connected external input. Choose a built-in source manually to keep using it. In Both mode, an external input supplies both recordings; Paired selects the separate built-in camera microphones.
The menu updates as devices connect or disconnect and shows the actual input during recording. Manual switches stay locked during a take. Routing is verified before reporting success, and repeated identical failure notifications do not trigger a switch loop.
41 automated camera, microphone, audio-session and paired-recording tests passed. Ed iPhone was unavailable, so a physical external-microphone recording was not verified.
Build 316 · 1.0
The microphone menu lists built-in sources reported by your device, with Auto as the default. Auto prefers the active camera’s direction when available.
Choose a source before recording. The control stays locked during a take; Both mode keeps its independent camera-matched microphones and displays Paired.
Source changes run off the main thread and check the active audio route. Camera lifecycle and audio-session tests passed in the simulator; physical microphone routing still needs device verification.
Build 315 · 1.0
Timeline scrubbing selects the clip under the playhead on the track you are working on. Select a clip on another track to switch the editing lane.
In a gap, the clip selection clears while the active track stays the same. Selection resumes when the playhead reaches another clip on that track.
The red playhead gesture deletes the selected segment and shifts later clips on that track left. Other tracks keep their timing. Links affected by this single-track move are detached, and Undo restores the edit and links.
Build 314 · 1.0
Tap a clip to select it, then tap it again for Split, Delete and More. All clip tools and effects are now in the menu in both timeline and clip views, freeing the space below the tracks.
Drag the playhead triangle toward the selected clip’s lower edge for a yellow split preview. Drag below the editor for a red delete preview. Release to apply, or move back up to cancel.
Playhead actions affect only the selected segment, including on multiple tracks. Drag deletion keeps the gap, other tracks retain their timing, and Undo restores the edit.
Build 313 · 1.0
Tap the live image to focus, with a brief yellow focus indicator. In Both mode, each preview focuses its own camera.
Focus works before and during recording while preserving exposure adjustments. Pinch zoom, exposure swipes, inset dragging and zoom buttons retain their existing actions.
Focus uses each camera’s supported autofocus capabilities and accounts for preview orientation, mirroring and letterboxing.
Build 312 · 1.0
Swipe a project card to the right to select it, or use the selection button. Tap cards to choose projects, with Select All, Deselect All and Done.
Duplicate selected projects or delete them with a confirmation showing the selected count. Batch changes are saved together and preserve media used by other projects.
Deleting the open project switches to a remaining project; deleting every project leaves a fresh empty project.
Build 311 · 1.0
Projects replaces the Settings tab, with automatic saving, rename, duplicate, delete, and video or audio export.
Open all app settings from the gear in Projects. Starting a new project now preserves the previous edit.
Reopen clips, tracks, effects, playhead position and export settings; protect shared media during storage cleanup.
Build 310 · 1.0
Reuse empty compatible tracks for camera recordings; create only the missing Main and Selfie tracks.
When entering Camera with existing footage, offer a single New Project action. Tap outside to keep the current project.
Keep recording and appending without repeated prompts while staying in Camera, with synchronized Main and Selfie tracks.
Build 309 · 1.0
Keep Open Gate recordings in the editor’s SDR color space instead of silently switching to HDR.
Preserve gradient precision through single-track and multitrack playback, with 10-bit opaque HEVC exports.
Scale camera recording bitrate with frame size and preserve 10-bit Both-camera capture when the camera supports it.
Build 308 · 1.0
Fix the selfie inset using stale camera rotation when entering Both mode or changing recording formats.
Drag the selfie preview with one finger; release to snap to a corner or an edge midpoint, keeping it inside the preview.
Keep pinch zoom and zoom buttons available, move the selfie controls with the inset, and retain its position across camera and format switches.
Build 307 · 1.0
Add a recording-format button next to the camera switch, with Standard and Open Gate choices based on the selected camera setup.
Capture supported native 4:3 or square formats without forcing a 16:9 output; preserve the same framing in preview and saved files, including the small selfie preview.
Lock format changes during recording, retain the choice when switching supported camera modes, and restore standard framing with an explanation if Open Gate cannot run.
Build 306 · 1.0
Discover built-in ultra-wide, TrueDepth and wide front cameras instead of taking the first wide-camera match.
Compare supported HD/30 fps selfie formats for wider views and native zoom options; apply the selected format and frame rate together before starting capture.
Keep zoom buttons tied to the active camera’s actual capabilities, retain the rear-camera selection, and try the next supported configuration if one cannot be configured.
Build 305 · 1.0
Merge overlapping and duplicate candidate moments during multi-part analysis before the existing final selection, including when Rearrange footage is on. Preserve adjacent cuts and Gemini’s final playback order.
Set the default analysis chunk duration to three minutes for AI Editing and AI Multicam. Existing saved chunk settings are preserved.
Include the failed response stage and conflicting numeric ranges in Copy error details. Final edits still reject invalid timestamps or repeated footage; no extra AI calls or automatic retries.
Build 304 · 1.0
Select two to four video tracks with stable Camera 1–4 labels and comparison layouts adapted to portrait, landscape or mixed footage.
Choose Follow camera, Mix all cameras equally, or keep one camera’s audio across every video switch. Fixed-source gaps stay silent.
Keep audio synchronized when trimming content and when combining the result into one track. Camera selections remain unique and audio choices follow the selected track when slots change.
Build 303 · 1.0
Use Camera 1 and Camera 2 consistently in track selectors, Gemini prompts, response schemas and comparison-video labels.
Map each camera number to its selected track for both video and audio, regardless of track names or recording roles.
Keep automatic Main/Selfie selection as a convenience for recordings; preserve actual track names and existing editing options.
Build 302 · 1.0
Add Trim content to AI Multicam. Each time it is enabled, target duration starts at zero (Auto), letting AI choose the length.
Choose kept moments and camera angles together in one request per analysis part. A nonzero duration is an approximate target; chronology is preserved.
Close removed time in video and audio together, including Mix both mics and combined or separate output tracks. Preserve unrelated tracks and restore the edit with Undo.
Build 301 · 1.0
Both mode shows arrows beside Main-camera zoom presets to switch between the device’s supported wider and longer lens groups before recording.
Keep zoom where the new group supports it, otherwise move to its nearest range endpoint. Preserve exposure compensation when switching groups.
Lens-group arrows are disabled during recording; pinch zoom and preset buttons continue working within the selected group.
Build 300 · 1.0
Compare supported camera and sensor-format combinations to select more native lens and sensor-crop zoom options in Both mode, without phone-specific settings.
Validate the complete two-camera and microphone setup against hardware and thermal budgets, with automatic fallback to a supported configuration.
Consider higher-resolution sensor formats when they add zoom options while keeping recorded video output at 1080p or below.
Build 299 · 1.0
Swipe vertically on the right half of the live camera image: upward brightens and downward darkens. Automatic exposure remains active with your chosen EV offset.
Use exposure compensation before or during recording, with a temporary EV readout and right-edge indicator. Adjust Main and Selfie independently in Both mode.
Keep pinch zoom and zoom preset buttons available, and apply exposure updates on the capture queue within the camera’s supported limits.
Build 298 · 1.0
Tap native zoom presets at the bottom of the camera feed. Available choices follow the active camera’s lens and sensor-crop zoom levels.
Preset taps switch immediately before recording and smoothly zoom to the selected level during recording. Pinching takes over and temporarily hides the preset buttons.
Keep Main and Selfie zoom controls independent in Both mode, with compact controls inside each image.
Build 297 · 1.0
Use two fingers to zoom while previewing or recording. The zoom multiplier appears near the bottom inside the camera image, then fades away.
Main camera uses the available virtual rear camera for optical lens switching. Both mode selects a lens combination supported by iOS; pinch the selfie inset to zoom it separately.
Keep zoom updates off the main thread and within the active camera’s supported range. No extra camera buttons.
Build 296 · 1.0
Add Merge Tracks to the track menu. Select tracks and render them into one clip, baking blending modes, grading, transforms and track audio effects.
Mix unmuted audio tracks equally after their effects. Preserve timing, undo and original media; keep master effects live. Audio-only selections produce an audio track.
Protect linked clips and layer order when selecting tracks.
Build 295 · 1.0
Add Mix both mics to AI Multicam, off by default. When enabled, both camera audio streams play continuously at 50% while the video switches.
Works with combined or separate tracks, preserving source timing, track mute, undo and export. AI analysis is unchanged.
Build 294 · 1.0
Format AI Multicam durations, chunk offsets and camera intervals as seconds with at most two decimal places.
Preserve precise timeline timing and resolve unambiguous rounded camera boundaries without weakening validation.
Build 293 · 1.0
Move Audio ducking below the Creative direction card in the AI audio dialog.
Show ducking as an expandable section, collapsed by default. Expand it to access the voiceover ducking toggle, amount and timing controls.
Build 292 · 1.0
Rename the clip effects menu entries and effect panel titles to Retro Camera and Time & Pitch.
Use the new names in processing and error messages. Preserve the original project credits and licenses in Open Source notices.
Build 291 · 1.0
Preserve unchanged waveforms when adding, rendering or editing other clips. Retain waveform data when thumbnail memory reaches its limit.
Display decoded waveforms before thumbnails finish, and retain them if thumbnail generation fails or is interrupted. Restart audio-card loading after rendering resumes or the cache resets.
Correct the bundled MP3 encoder’s integer conversions, unsigned bit masks and comparison warnings without disabling diagnostics. Normal MP3 output is byte-identical at all supported bitrates.
Build 290 · 1.0
Add 64, 96, 128, 160, 192, 256 and 320 kbps export choices for MP3 and M4A. Preserve the existing defaults of 320 kbps for MP3 and 256 kbps for M4A.
Add 16-bit and 24-bit WAV choices, with the corresponding uncompressed bitrate displayed. The default remains 24-bit.
Apply quality selections to the encoders and export cache, while retaining the complete audio mix and normalization settings.
Build 289 · 1.0
Remove the small add-track plus button beneath Add Media in the empty editor.
Choose Video or Audio in export settings. Audio-only projects export audio directly; MP3 is the default audio format, with M4A and WAV also available.
Audio exports include the complete clip, track and master mix and the chosen normalization setting. Changing format correctly selects a new export instead of reusing another format.
Encode MP3 at 320 kbps, AAC in M4A at 256 kbps, and WAV as 24-bit PCM, all at 48 kHz stereo. Include the dynamically linked LAME encoder license and source in Open Source settings.
Build 288 · 1.0
Separate audio now names the tracks from their source track, such as Track 1 Video and Track 1 Audio, including custom track names.
The Audio page hides populated tracks whose clips no longer contribute audio, while retaining tracks that still contain audio, muted audio tracks, and empty audio-capable tracks.
Track naming remains undoable and saved with the project. Repeated separation does not repeat the Video suffix.
Build 287 · 1.0
Choose Video & Audio, Video only, or Audio only when adding a track. Import audio files and see audio clips as waveforms.
Separate a clip's audio without copying or re-encoding its source. Linked position, trim and split edits preserve synchronization; Unlink allows independent editing.
Route picture and audio by track type through playback, effects and export. Audio-only renders bake track effects once, and audio-only projects export WAV with optional normalization.
Preserve undo, saved projects, clip gaps and cross-track editing, including linked-move collision checks.
Build 286 · 1.0
Expand all three live equalizer gain controls from ±12 dB to ±24 dB.
Use the same range for the controls and audio-effect validation, including saved settings and rendering.
Keep the existing 0 dB reset and 0.5 dB slider steps.
Build 285 · 1.0
Process audio in the order: optional mono conversion, optional low cut, denoising, then the Strength mix.
Strength now blends the low-cut original with its denoised copy. No low cut is applied after the mix.
Changing low cut clears the prepared preview so the next Preview request uses the correct model input. Defaults remain enabled at 100 Hz with a 20–300 Hz range.
Build 284 · 1.0
Add Low cut toggle and a 20–300 Hz frequency slider to Voice Cleanup. The default is enabled at 100 Hz, restored by Reset controls.
Apply a gentle 12 dB/octave high-pass to the final cleanup mix. Preview and Apply use the same filter; Original comparison bypasses it.
Release temporary ONNX inference objects after every audio processing step instead of retaining them throughout cleanup.
Preserve the existing denoising model, sound, timing, mono conversion and manual preview behavior.
Add sustained inference and full mono/stereo processing memory regressions alongside playback, export, undo and audio-reference checks.
Build 282 · 1.0
Opening Voice Cleanup or returning to the app no longer starts audio processing. Tap Preview to generate and play a sample, or Apply to process without previewing.
Add Convert to mono, enabled by default and restored by Reset controls. Stereo channels are averaged before denoising; disable it to keep independent stereo processing.
Changing controls pauses the preview and waits for Preview again. Changing the mono option clears the previous processed sample.
Verify pre-inference mono conversion, stereo opt-out, manual preview generation, direct Apply, playback, export, undo and persistence.
Build 281 · 1.0
Add Voice Cleanup to the selected clip menu using the bundled DPDFNet8 48 kHz HR model. Processing stays on the device.
Preview a short sample, compare the original audio, adjust strength, and apply in one undoable edit. Reset controls includes haptic feedback.
Process mono or stereo with independent channel states and compensated model delay. Retain lossless intermediate audio, source video quality, clip timing, grade and track effects.
Add Apache 2.0 licensing and model attribution to Open Source notices.
Verified against the upstream Python reference, including speech boundaries and independent stereo channels. Physical iPhone tests cover preview switching, apply, playback, export, cancellation, undo and persistence.
Build 280 · 1.0
Add Combine into one track, enabled by default. Edited shots remain individually editable on a Multi-Cam track; turn it off to keep Main and Selfie separate.
Preserve source trims, timing, gaps, clip grades and audio-follow-picture cuts in one undoable edit. The combined track retains Main’s track-level audio effects, with a note shown when source effects exist. Unrelated tracks are preserved.
Match AI Editing’s large full-width Create edit button, confirmation buttons and progress styling.
Verified combined and separate edits through playback, export, track mute and undo on Ed iPhone; verified source audio switches at camera cuts. Simulator tests cover timing, gaps, clip settings, unrelated tracks and persistence.
Build 279 · 1.0
Restore rear-facing audio to Main and screen-facing audio to Selfie, undoing the microphone swap in build 278.
Applies to new dual-camera recordings. Existing recordings retain their embedded audio. Gemini continues receiving the 50/50 audio mix.
Build 278 · 1.0
Swap the two directional microphone feeds during capture to address the reported reversed Main/Selfie audio. Applies to new recordings; existing files retain their embedded audio.
Preserve synchronized recordings, the 50/50 Gemini analysis mix, and audio that follows the chosen camera during AI multicam edits.
All nine targeted camera writer, frame pairing, AI source mapping and persistence tests passed. Physical microphone direction still needs verification on the iPhone.
Build 277 · 1.0
Allow AVFoundation to negotiate front/back directional microphone settings for dual capture while retaining serialized audio-session ownership and built-in microphone selection.
Clean up partially activated audio sessions on setup failure so Retry and Main camera can recover. Preserve actual runtime errors and identify missing startup streams instead of a generic camera timeout.
20 targeted camera lifecycle, audio-session, pairing and separate audio-writer tests passed. Physical camera startup could not be verified because the iPhone remained unavailable.
Build 276 · 1.0
Record Main and Selfie with separate built-in rear/front directional microphone audio on the same clock. Wait for both audio streams before enabling recording.
Mix both source microphones at 50% each only for the numbered Gemini comparison upload. Camera switches apply to both picture and its matching audio, with existing tiny anti-click fades and undo.
Updated Gemini instructions to favor natural speaking boundaries. Tested separate encoded audio streams, synchronized timing, 50/50 proxy levels, camera-cut sound routing, state/undo, and a live Gemini request.
Physical directional microphone pickup and iPhone playback remain unverified because the iPhone was unavailable. Simulator audio-device limitations prevented two playback/audio-tap checks.
Build 275 · 1.0
Add AI Multicam beside AI Editing: choose two tracks, analyze a numbered side-by-side or stacked comparison, and apply undoable camera switches without shifting time.
Automatically select Main and Selfie camera tracks; retain continuous Main audio and mute Selfie by default, with track mute controls in Audio and track menus.
Share the configurable AI Editing chunk duration, carry context between parts, and validate Gemini's camera choices before changing the project. Tested live requests with portrait and landscape fixtures.
Keep continuous sound in exports, track effects, saved projects and undo. Physical iPhone playback testing was unavailable because the device was disconnected.
Build 274 · 1.0
Move playback audio activation off the main thread and serialize it with camera audio ownership, preventing delayed previews from interrupting capture.
Rebuild the combined timeline when returning from Camera after another recording, even if the previous selected clip remains loaded.
Fix the reported Swift concurrency, compositor, shader and Sonic compiler warnings. Verified three consecutive dual-camera recording, import and playback cycles on a connected iPhone.
Build 272 · 1.0
Both mode now receives Main and Selfie frames independently and pairs nearby exposures on the shared capture clock, instead of requiring identical timestamps in one callback. Frame buffering stays bounded when either camera stalls.
Record becomes available after both cameras deliver usable frames. Empty failed takes are cleared automatically, while recordings containing data are kept for recovery.
Improved readability of the Add to timeline recovery button.
Build 271 · 1.0
The camera switch now cycles through Main, Selfie, and Both on devices supporting simultaneous front and rear capture. Both mode shows a selfie inset over the main camera preview.
Dual recordings become separate synchronized Main and Selfie tracks. Later takes reuse those tracks and append at a shared start time. Microphone audio is recorded once, on Main, to avoid doubled sound.
Both files are saved together with retry protection and a single undo step. Recording remains on the Camera tab. Devices without dual-camera support retain Main and Selfie modes.
Camera startup and shutdown are serialized so switching away or backgrounding during dual-camera setup cannot leave a late capture session running.
Build 270 · 1.0
The camera switch now cycles through Main, Selfie, and Both on devices supporting simultaneous front and rear capture. Both mode shows a selfie inset over the main camera preview.
Dual recordings become separate synchronized Main and Selfie tracks. Later takes reuse those tracks and append at a shared start time. Microphone audio is recorded once, on Main, to avoid doubled sound.
Both files are saved together with retry protection and a single undo step. Recording remains on the Camera tab. Devices without dual-camera support retain Main and Selfie modes.
Build 269 · 1.0
Stopping a recording keeps the Camera tab and live preview open. Finished videos are added in the background to the end of the selected track, including after any existing gaps.
Camera imports preserve the current edit selection and avoid loading the recorded video into the editor's player. The camera is ready for another recording once saving finishes.
Build 268 · 1.0
Camera replaces Seedboard with an embedded live preview, front/back camera switching, and video recording with sound. Stopping adds the recording to the selected timeline track without saving to Photos.
Recording files stay in app storage until timeline import succeeds. Camera interruptions stop recording, and failed imports can be retried.
Removed Seedboard-specific views, state, generation tasks, and preferences. Shared grading, randomization, presets, and editing tools remain.
Build 267 · 1.0
Timeline scrubbing completes one video seek at a time, then jumps to your latest finger position instead of repeatedly cancelling the decoder.
Preview quality and accurate final positioning are unchanged. Playback can resume after the final seek completes.
Build 266 · 1.0
Add Volume & Pan to any audio track or Master: volume from −24 to +24 dB, left/right pan, and stereo width from mono to 200%.
Live controls start at 0 dB, centered pan and 100% width. Supports repeated instances, reordering, reset, removal and Undo, and is included in playback and exports.
Build 265 · 1.0
Normalize audio appears under Include audio and is on by default. It adjusts the completed soundtrack after track and Master effects.
Normalization uses one uniform gain with no boost cap, leaves silence silent, and does not change the project’s mix or effect settings.
Build 264 · 1.0
Switch between Tracks and Master beside the audio playback controls. Add, reorder and adjust independent effects on the combined output.
Master supports EQ, Echo, Reverb, Compressor, live metering, measured compressor gain and Undo. Settings are saved with the project and included in playback and final exports.
Build 263 · 1.0
EQ, Reverb, Compressor and Echo controls stay available when off. Adjusting a slider automatically enables that effect.
Compressor Output and Analyze & Set Gain now support up to +48 dB. Gain analysis can also enable a bypassed compressor when applying its result.
Build 262 · 1.0
Compressor Output now ranges from −12 to +24 dB. Removed automatic make-up gain.
Analyze & Set Gain measures the whole track through the selected compressor and sets Output for a −0.1 dBFS sample peak, within slider limits. Analysis is cancelable and the adjustment supports Undo.
Build 261 · 1.0
Camera videos containing stereo and Spatial Audio alternatives now use the enabled recording instead of mixing both together.
Applied consistently to preview, export, scrubbing and audio-processing paths while preserving independent enabled audio tracks.
Build 260 · 1.0
Waveform bars now follow separate short slices of each track’s processed audio, instead of all sharing one smoothed average and a fixed shape.
Loud output has more display headroom. Centered animation, the existing color palette, clipping warning and static silent/paused icon are preserved.
Build 259 · 1.0
Tap a track’s waveform or name to collapse its effect controls and tap again to reopen the last viewed effect. Effect buttons stay visible.
Reset and Remove now share one row beneath every effect. Removing an effect still asks for confirmation.
Build 258 · 1.0
Audio track waveform bars now grow equally upward and downward from their center as the processed audio level changes.
The existing blue-to-teal-to-gold colors and red clipping warning remain. Paused or silent tracks show the original static waveform icon.
Build 257 · 1.0
Track waveform icons keep their shape and smoothly change from muted blue to teal to gold with the processed audio level. Silence returns to dim gray.
Clipping at a track’s output briefly turns its icon red for 1.5 seconds. Readings follow playback time and stay separate for each track.
Build 256 · 1.0
Empty audio effects now appear as faint rounded buttons with a small centered plus, matching the size of effect buttons.
Start with four slots; adding effects replaces placeholders and always leaves an add slot available. Longer chains remain horizontally scrollable.
Build 255 · 1.0
Empty tracks show subtle + slots. Add EQ, Echo, Reverb or Compressor, including multiple instances with independent settings.
Scroll longer chains horizontally and hold-drag to reorder, with edge scrolling. Remove an effect from its controls or drag it outside the track card, then confirm.
Existing effect settings and order are preserved. Effect chains support undo, playback, export and rendering into clips.
Build 254 · 1.0
The default audio effect order is now EQ, Echo, Reverb, Compressor.
Tracks with a saved custom effect order keep that arrangement.
Build 253 · 1.0
Moved the effect buttons and flow arrows slightly upward, balancing the gap below the track heading with the space below the arrows.
The audio track cards keep their existing size.
Build 252 · 1.0
Thin arrows beneath each track’s audio effect buttons show the processing order between enabled effects, skipping effects that are turned off.
The arrows update when effects are toggled or reordered and fit within the existing card size.
Build 251 · 1.0
Settings now lives in the main tab bar instead of opening as a modal sheet. The bottom tabs stay available while browsing settings.
Removed the sheet’s Done button. Settings subpages keep their native back navigation, and existing controls and saved values are preserved.
Build 250 · 1.0
Added Echo with Time, Feedback and Mix controls to every track in Audio.
Tap Echo for its controls or hold and drag it to change processing order. Green dots show enabled effects; the white button shows the open controls.
Echo settings support reset, undo, saved projects, playback, export and track rendering. Existing projects keep their effect order and start with Echo off.
Build 249 · 1.0
The white background now marks the effect whose controls are open. Switching or closing the controls updates that highlight.
The fixed left status dot independently shows whether each effect is enabled: green for on and gray for off.
Build 248 · 1.0
EQ, Reverb and Compressor status dots now use a fixed left position inside each button, independent of the label length.
Build 247 · 1.0
Added a Compressor to each track in Audio, with threshold, ratio, attack, release, output gain and an Auto gain switch.
Tap Comp to open its controls. Hold and drag to set its processing order alongside EQ and Reverb.
Compression uses a linked stereo detector and a soft knee. Auto gain is on by default; the compressor starts bypassed so existing projects sound unchanged.
Settings and order are saved per track and support undo, reset, playback, export, Render Clips and Merge All.
Build 246 · 1.0
Touch and hold EQ or Reverb, then drag left or right to rearrange that track’s effects. The audio processing order follows the buttons.
Removed hold-to-toggle. Tap an effect for its controls and use its switch to turn it on or off.
Effect order is saved per track, supports undo/redo, and is used for playback, scrubbing, export, Render Clips and Merge All.
Dropping outside the effect row cancels the move. Reordering preserves the selected controls and each effect’s settings.
Build 245 · 1.0
Added live track-wide reverb with Mix, Decay, Pre-delay, and Damping controls. Reverb starts off and can work alongside EQ.
Tap EQ or Reverb to switch the controls shown inside that track card. Hold either effect button to toggle it, with its own status dot and haptic feedback.
Track effects persist with the project, support undo and reset, and are included in exports and Render Clips / Merge All.
Fixed effect-only undo rebuilding playback in older single-track projects.
Build 244 · 1.0
Added a status dot inside the EQ button: green when enabled and muted when off.
The four-slot layout and tap/hold controls remain unchanged.
Build 243 · 1.0
Audio cards now reserve four equal-width effect slots, with a compact EQ button and three empty spaces for future effects.
Tap EQ to show or hide its controls. Touch and hold to toggle the effect without changing its settings, with haptic feedback.
The button shows whether EQ is enabled. Hold and tap actions are exclusive, and toggling supports undo and project saving.
Build 242 · 1.0
Moved each track name onto the same row as its wave icon in the Audio panel, reducing card height.
Build 241 · 1.0
Replaced the Looks tab with Audio, with one expandable card per editor track and Bass, Mid, and Treble equalizer controls.
Equalizer applies to the track's original audio, narration, and music together, with bypass, reset, project saving, and undo.
Playback and export use the track EQ; Render Clips and Merge All bake it once and clear the live effect.
Verified targeted audio/playback regressions and native Audio-screen interactions on iPhone and iPad simulators.
Build 240 · 1.0
Prepared preview audio now ends at the exact video timeline duration, preventing sub-sample rounding from leaving playback without video frames.
Preserves the existing audio mix, tiny anti-click fades, clip positions, and live grading.
Verified six native cross-track drag gestures, five seconds of playback after each move, and playback after relaunch on the iOS 26.4 simulator.
Build 239 · 1.0
Moved sequence still-image readback off the UI thread, with one conversion at a time and stale-result rejection.
Moved paused source-preview composition setup and image-generator cancellation off the UI thread, rejecting work made obsolete by a newer seek.
Placement-only clip moves now reuse the selected source decoder and grading state while rebuilding the actual timeline composition.
Added timeouts and cancellation to source-preview and clip-thumbnail frame requests, so a missing media callback cannot leave those requests waiting indefinitely.
Build 238 · 1.0
Isolated native drag sessions so delayed callbacks cannot interrupt a newer drag or commit its edit early.
Used compact lift and cancel previews instead of snapshotting the full-width timeline clip.
Prevented scroll configuration callbacks from recursively changing timeline state.
Playback health now recognizes advancing decoded frames even when they briefly lag the audio clock. Frozen video detection remains enabled.
Removed an unnecessary GPU still-image capture when replacing the preview surface.
Build 237 · 1.0
Fixed a missing native drag-completion callback that could leave clip dragging active and timeline scrolling locked after a drag ended without destination cleanup.
Completed moves still commit once; cancelled or forbidden drags discard pending changes and restore scrolling.
Build 236 · 1.0
Removed the 16-point left, right and bottom padding from the timeline view.
Expanded the usable track area while retaining empty-space scrubbing for single and multiple tracks.
Build 235 · 1.0
The single-track timeline now uses the available track area, allowing horizontal scrubbing from the empty space beneath the clips, just like multiple tracks.
Timeline padding remains 16 points on the left, right and bottom.
Build 234 · 1.0
Changed the default full-screen Tint sensitivity to 0.25×, matching Temperature / White balance.
Reset full-screen controls and Reset Settings use the new default. Existing saved sensitivity choices remain unchanged.
Build 233 · 1.0
Same-track gaps now accept dragged clips at the indicated time, including gaps left by Keep Gap deletion.
The drop marker and final placement snap to the playhead within a narrow 8-point radius that stays consistent as the timeline is zoomed.
Tap the selected clip again to deselect it and hide its controls in both timeline and clip views. Tapping another clip selects it normally.
Build 232 · 1.0
Dragging the empty space below the timeline now scrubs every track together; tapping still deselects.
Drag a clip onto another track to place it at the drop time. Its grade, audio and effects stay intact. The source keeps its gap, and overlapping destination clips move later without being overwritten.
In clip view, track drop targets appear while dragging and preserve the clip’s existing start time. Cross-track moves are undoable and prepare the updated playback composition before returning control.
Build 231 · 1.0
In edits with multiple tracks, tapping empty space clears the clip selection in timeline and clip views.
The selected clip controls disappear, and timeline mode expands the track area into the freed space.
Deselecting preserves the playhead, composed preview and active track. Playback and multitrack scrubbing keep the selection empty until another clip is selected.
Build 230 · 1.0
Moved White balance & tint gestures into Full-screen controls.
Added Reset full-screen controls with Cancel/Reset confirmation and haptic feedback. It restores only this page’s settings.
Relative adjustment is now the default. Sensitivity defaults: Gain, Saturation and Contrast 0.5×; Temperature/White balance 0.25×; Tint 0.1×. Saved choices remain unchanged until reset.
Build 229 · 1.0
Full-screen controls now offers Range limits or Relative adjustment. Relative swipes add to or subtract from each current grading value, within the regular slider limits.
Relative mode shows Sensitivity for Gain, Saturation, Contrast, Temperature and Tint instead of Minimum and Maximum. Repeated swipes continue from the previous value.
Ranges and sensitivities are saved independently when switching modes. Existing range mode remains the default. Reset Settings restores range mode and 1× sensitivities.
Build 228 · 1.0
In fullscreen, two fingers immediately reveal the original for as long as both stay down. Releasing either finger restores the grade. Removed one-finger hold comparison and two-finger undo.
Settings adds Full-screen controls with saved minimum and maximum values for Gain, Saturation, Contrast, Temperature and Tint. Existing fullscreen limits remain the defaults; Reset Settings restores them.
Regular preview comparison and grading sliders are unchanged. Tap playback, double-tap randomization, three-finger reset and fullscreen swipe controls remain available.
Build 227 · 1.0
Settings adds White balance & tint gestures, on by default. Horizontal swipes in the top third adjust Tint; the middle third adjusts Temperature (white balance); the bottom third scrubs playback.
Turning the setting off restores horizontal scrubbing anywhere. Existing vertical Gain, Saturation and Contrast gestures keep their 40% / 20% / 40% regions.
Gestures keep the control chosen at touch-down, show temporary adjustment feedback, and support undo. Reset Settings restores the new toggle to on.
Build 226 · 1.0
Removed the fullscreen eye button, playback panel, seek slider, time display, Randomize button, loop toggle, comparison button and Original/Graded label.
Fullscreen gestures, A/B/C look slots, exit button and temporary scrub/adjustment indicators remain. The scrub indicator now aligns beside the exit button.
Build 225 · 1.0
Rearranging clips preserves their existing waveform and thumbnail previews instead of showing a false silence line.
Preview completion now refreshes the timeline directly. Changes to clip ranges or audio still regenerate the appropriate preview.
Build 224 · 1.0
Dragging a clip leaves a faded, desaturated placeholder in its original slot in both timeline and clip views.
Split pieces use underscore names such as Clip 1_1, Clip 1_2 and Clip 1_3. Further splits keep numbering flat; rearranging preserves each piece's name.
Settings now includes Resume playback after scrubbing, off by default. When enabled, dragging the regular timeline during playback resumes from the release position. Previously paused playback stays paused.
Build 223 · 1.0
Edited playback now verifies its source files, cut ranges, order and track placement before reusing a player. Interrupted preparation cannot leave the previous uncut timeline active.
Dragging clips now restores timeline scrolling through both source and destination drag-completion paths, including when the source cell scrolls offscreen.
Render Clips is disabled when nothing needs baking. On mixed tracks, it renders only clips with pending grading, AI audio, mute, trims, photos or frame/corner changes.
Build 222 · 1.0
Added pitch-preserving Sonic speed fitting for the combined AI narration after existing pause and timing adjustments, before repeating footage. Music is unchanged.
The narration uses one consistent speed, choosing the smallest increase that fits when possible. Any overrun remaining at the limit uses the existing footage extension.
Added Maximum narration speed in Settings → AI audio: 1.3× by default, adjustable from 1.0× to 2.0×. Set 1.0× to disable. Applies to new generations; Reset Settings restores 1.3×.
Updated narration placement and speech-ducking analysis for the fitted audio. Resuming a completed fitting stage does not process it again.
Build 221 · 1.0
Horizontal dragging now scrubs in the opposite direction in both regular and full-screen previews: drag right to advance and left to go back.
Added Settings → Preview scrubbing → Reverse scrub direction. It is off by default; turn it on to restore the previous direction in both previews. Your choice is saved, and Reset Settings turns it off.
Build 220 · 1.0
Added Sonic Audio beneath PowerShot Effect in each video clip’s menu, with Speed, Pitch, Tape rate, Volume and Standard/High quality controls. Standard quality is selected by default.
Preview a short excerpt in the fixed video window, then apply the effect to the selected clip. Video follows speed and tape-rate changes; grades and transform controls remain editable. Undo restores the original clip.
Processing runs locally using Sonic’s Apache-licensed audio core. No spectrogram or FFT dependencies are included. Added its license to Open-source licenses.
Build 219 · 1.0
Added Language-aware timing in Settings → AI audio → API keys, voices & generation. Enable it to ask Gemini to estimate spoken length using the selected language’s pronunciation and delivery style.
When disabled, the original narration instruction is unchanged. The option applies to new generations and is saved with each job so all chunks and resumed requests use the same mode.
Reset Settings turns the option off. Existing scripts, generated audio, duration cushions and audio-fitting behavior remain unchanged.
Build 218 · 1.0
Tapping Reset Settings opens a native iOS confirmation with Cancel and Reset buttons. Settings change and haptic feedback occurs only after choosing Reset.
Cancel leaves all controls unchanged. Projects, saved looks, media and API keys remain preserved.
Build 217 · 1.0
Added Reset Settings in the app Settings page. It restores AI Audio and AI Editing preferences, fullscreen scrubbing, pause-removal defaults, Seedboard controls, viewer controls, loop playback and randomize-on-import.
Projects, clip grades and soundtracks, saved presets, A/B/C looks, startup looks, media and API keys are preserved.
Added soft haptic feedback to AI Audio reset, VHS and PowerShot reset controls, Speed & Reverse reset, photo duration reset, Restore Full Clip and transform reset. Existing grading resets keep their feedback without duplicate pulses.
Reset Settings and the successful Saved Looks & Settings reset in Storage also give the same soft haptic.
Build 216 · 1.0
Speech across chunks can now start an overlong narration in the previous chunk’s unused tail, moving it only as early as needed and leaving a short gap after the previous voice. The same toggle still allows speech to continue into the following chunk.
Timing accounts for earlier spillovers so narrations do not overlap. Neighboring space is used before pause compression and the existing repeated-footage fallback, and ducking follows the final assembled narration.
The default AI audio chunk duration is now 20 seconds. Saved custom durations and already generated soundtracks are preserved.
Build 215 · 1.0
Swipe left on Project Files, Unused Files, Exports, Temporary Files or Saved Looks & Settings to reveal a red trash action. Tapping it asks for confirmation before deletion.
Removed the separate cleanup buttons below the categories. Clear All Media is available in the top-right Storage options menu.
Saved Looks & Settings removes saved presets and comparison looks and resets preferences while preserving the project, its grades, media and API keys.
Build 214 · 1.0
New soundtracks default to 50% music and enabled ducking at 75%, with 500 ms attack, 500 ms release, 200 ms hold and 250 ms lookahead.
A Reset button at the bottom of AI audio restores soundtrack mix, ducking and creative direction defaults while retaining generated audio and scripts.
Reset restores Original 3%, Voiceover 100%, Low Cut 100 Hz, Epic trailer, Enceladus, English, Roast mode off and an empty direction field. Previously saved custom settings remain until changed or reset.
Build 213 · 1.0
Audio ducking now has adjustable Attack, Release, Hold and Lookahead sliders, grouped with the enable switch and amount between divider lines.
Timing settings are saved with each clip and used by preview, rendered clips and export. Choices are also remembered for new soundtracks.
Existing projects retain the previous default ducking timing: 80 ms attack, 250 ms release, 80 ms hold and 80 ms lookahead.
Build 212 · 1.0
Cached preview covers are now tied to their playback position so an earlier frame cannot reappear during playback preparation.
Fullscreen pause keeps the composed timeline player in place instead of switching to a clip source player.
Starting playback during an unfinished scrub explicitly completes the requested seek before preroll.
Build 211 · 1.0
Rearranging clips now prepares a new composed timeline at the playhead instead of retaining playback for the old clip order.
Drag scrolling starts only within 24 points of the left or right edge, with a capped speed. Normal timeline scrolling is restored when the drag ends.
Drops commit after the drag has finished. Trailing taps cannot jump to a selected clip, and the playhead position from the start of the drag is preserved.
Build 210 · 1.0
Press and hold a clip, then drag it before or after another clip in the same track. Works in both clip and timeline views.
A drop marker shows the insertion position. Each completed move is undoable and preserves the clips' grades, audio, effects, and track gap lengths.
Dragging pauses playback; cancelling leaves the edit unchanged. Normal tapping and timeline scrubbing keep their existing behavior.
Build 209 · 1.0
Horizontal dragging on the regular preview scrubs only the selected clip, in both timeline and clip views.
Scrubbing pauses playback, respects trimmed clip boundaries, and keeps the current clip selected. Fullscreen scrubbing continues to use the whole timeline.
Holding without dragging shows the original; moving or resizing a clip keeps its transform gestures.
Build 208 · 1.0
Deleting a track now prepares playback for the remaining timeline, including when two tracks become one.
Selection moves to surviving footage instead of getting stuck on a deleted clip or empty neighboring track. The playhead is preserved within the remaining duration.
Pending playback and recovery work is cancelled before track changes; controls become available after the replacement preview is ready.
Build 207 · 1.0
Selecting a clip in clip view now moves the playhead and preview to that clip’s start. Timeline-view taps continue to select without moving the playhead.
Added Paste Grade to All Clips to the track options menu in both layouts. Copy a grade first, then apply it to every clip on the chosen track.
Track-wide grade paste preserves timing and other tracks, and can be undone in one step.
Build 206 · 1.0
Restored projects and ordinary timeline starts now use the same prepared opening-frame path as newly completed AI edits.
Every new timeline player gets its own correctly positioned composed frame instead of reusing a source or previous preview image.
Verified a wrong-frame regression and first playback through three saved-project reloads with reordered iPhone HEVC footage.
Build 205 · 1.0
The edited preview now prepares its actual video decoder, audio, and opening frame before it is ready to play.
Starting at the prepared position preserves decoded frames instead of seeking again. Still-image conversion no longer runs when playback starts.
Checked cold first playback on H.264 and iPhone HEVC footage, reordered cuts, fullscreen return, looping, and opening narration.
Build 204 · 1.0
Fullscreen and the studio now use one persistent video display layer. Exiting fullscreen preserves position without competing video outputs.
Adjacent cuts alternate across reusable video decoder lanes to reduce stalls at source jumps. Recovery now detects frozen video even if audio keeps playing.
Rearrange footage explicitly requests a changed story order. Chronological results are rejected rather than silently presented as rearranged; no automatic API retries were added.
Verified with live single-part and multipart Gemini requests, first-start frame checks on H.264 and iPhone HEVC footage, fullscreen exit, export of the live AI result, undo, looping, and audio-fade regressions.
Build 203 · 1.0
Returning from fullscreen preserves the timeline position instead of reopening the selected clip at its start.
Fullscreen and the regular preview share the current rendered frame, preserving its orientation during the handoff and next playback start.
Verified both reproduced failures and nine focused simulator checks covering AI-edited cuts, fullscreen exit, looping, playback, and scrubbing.
Build 202 · 1.0
The timeline's fallback image now refreshes after paused seeks, so playback preparation cannot bring back a cached shot from an earlier AI cut.
Frame capture checks the current player item and frame time. Refreshing is limited to seek and readiness changes to avoid continuous still rendering.
Verified with an AI-edit regression that failed at two cut boundaries before the fix and passes afterward, plus normal and fullscreen playback checks.
Build 201 · 1.0
Scrubbing a single-track edit automatically selects the clip under the playhead, including edits with gaps. Multiple-track scrubbing keeps the chosen clip selected.
Tapping a clip in timeline or clip view now selects it without jumping playback. The selection outline remains visible even away from the playhead.
The composed preview stays attached during selection, and the grading source updates after scrubbing finishes.
Build 200 · 1.0
Timeline playback now keeps the correctly positioned video layer visible while preparing to play, instead of briefly restoring a cached frame from another point.
The fix also covers resuming an edited sequence after seeking, while preserving the existing still cover for unready video and single-clip playback.
Build 199 · 1.0
Tapping Create edit replaces it with equal-width Cancel and Confirm buttons, without a separate confirmation pop-up or new accent colors.
Cancel restores Create edit without starting a request; Confirm starts generation.
Build 198 · 1.0
The target clip count displays a compact estimate such as ≈ 15 · 2 s/clip, updating as either target length or clip count changes.
The average uses the requested final edit length. Auto stays unchanged, with no additional explanatory text.
Build 197 · 1.0
Added Settings → Storage with total usage, file categories, individual file details and previews, and confirmed deletion.
Clear Temporary Files, Delete Unused Files, or Clear All Media. Saved looks, preferences, API keys, and originals in Photos or Files are kept.
Cleanup protects active work and saved dependencies, removes affected clips safely, rebuilds playback caches, and recovers interrupted deletions.
Build 196 · 1.0
Speech across chunks is on by default in AI audio settings: narration can use spare room in the next chunk before pause compression or repeated footage is needed.
Voiceover ducking is on by default in AI Audio. Adjust its amount to lower original sound and music during speech, with smooth recovery between phrases.
Ducking settings stay with the clip and apply consistently to preview and export.
Build 195 · 1.0
During playback in clip view, the clip under the playhead becomes selected and scrolls into view automatically.
Selection follows cuts and fullscreen loops without replacing or pausing the running sequence player.
Pausing returns editing controls to the current clip. Timeline view keeps its independent selection.
Build 194 · 1.0
Deleting the selected clip in clip view now loads the remaining selection directly, so Play no longer depends on tapping another clip first.
Play repairs an unloaded or stale clip focus before starting, instead of using a previous clip’s playback item.
Timeline-mode deletion keeps its existing playhead and gap behavior.
Build 193 · 1.0
Added Rearrange footage, off by default, to let Gemini choose a more engaging playback order. The chosen order is preserved through preview and export.
Added Target clip count, from Auto (zero) to 60. Positive values are approximate targets for the whole edit; Auto sends no clip-count instruction.
Both preferences are remembered. Adjacent chosen ranges remain separate editable clips, and the result remains undoable. AI Editing requests still have no automatic retries.
Build 192 · 1.0
The export button immediately reopens Save/Share for the most recent completed export when the edit, source files and export settings are unchanged.
Changes to picture, cuts, tracks, audio mix or export settings require a fresh render. Missing cached files are regenerated; cancelled or failed renders never replace a completed export.
Touch and hold the export button to open export settings when a completed file is available.
Build 191 · 1.0
Clip playback silently retries twice, rebuilding the playback item while preserving position, grading and audio mix.
Stopping playback, leaving full screen or putting the app in the background cancels pending recovery so it cannot restart playback or show a stale error.
Timeline recovery remains bounded. AI Editing API requests still have no automatic retries.
Build 189 · 1.0
AI Editing now constrains cut times to the actual clip duration and includes a visible elapsed-time reference only in the temporary analysis video.
Empty placeholders and overlapping keep ranges are handled without duplicating footage. Cuts outside the video remain rejected.
Removed automatic retries from AI Editing. Failed requests are shown immediately, with copyable validation and provider details.
Build 188 · 1.0
Added AI Editing with an approximate target length and optional direction. Gemini chooses moments from the selected clip while keeping their original order.
Analysis uses a compact 15 fps video copy with audio. Longer clips are reviewed in configurable chunks, defaulting to five minutes, before one final selection.
Edits preserve original files, grades, framing and audio settings, with tiny anti-click fades. Undo restores the previous edit.
Added AI Editing analysis settings and progress showing each upload size.
Build 187 · 1.0
Randomization in Edit, Grade, module controls and full-screen double-tap now gives a short soft haptic.
Added the same feedback to grade resets, including long-press reset, full-screen reset and module resets.
Previous and next preset arrows also give a short soft haptic.
Build 186 · 1.0
Added a short soft haptic to randomization in Edit, Grade, module controls and full-screen double-tap.
Previous and next preset buttons now give the same haptic feedback.
Build 185 · 1.0
Added Frame & Corners to the clip menu: live corner radius, border width, white-default color picker, and Inside / Center / Outside placement.
Adjust against the actual edit preview, then Cancel or Apply as one undoable change. Reopen anytime to readjust.
Preview and export follow clip transforms and clip everything to the canvas. Render Clips and Merge All bake framed overlays with transparency and reset the baked settings.
Removed two-finger randomization from the regular Edit and Grade previews. Full-screen gestures are unchanged.
Build 184 · 1.0
In Transform mode, double-tap the preview to snap rotation to the nearest 45 degrees and position to the nearest canvas edge or center alignment.
Scale stays unchanged. A short soft haptic confirms the snap, and the complete alignment can be undone in one step.
Dragging, resizing and rotating remain free between snaps.
Build 183 · 1.0
Moved previous and next preset buttons out of Save Current Look and into the heading row.
Both buttons match the A/B/C button height, use twice their width, and keep the same spacing between buttons.
Build 182 · 1.0
Added Presets immediately before Main grading in the grading module row.
Use the same saved looks, apply, save, import, share and manage presets without leaving Grade. The Looks tab remains available.
Entering Grade selects Main grading by default.
Build 181 · 1.0
Opening the PowerShot effect now selects Film with its default film stock.
Film starts with exposure, brightness and contrast at zero. Other signal modes remain available.
PowerShot now defaults to Exposure −2, Brightness −0.25 and processing resolution 1440 px.
Build 175 · 1.0
PowerShot now defaults to Exposure −1.5, Brightness −0.1 and Contrast 1.
Overall mix is the first slider below the preview and blends the complete effect, including exposure, brightness and contrast.
Build 174 · 1.0
Fixed the component-order typo in the PowerShot temporal-noise hash to match the original implementation.
Build 173 · 1.0
Added PowerShot Effect below VHS in the clip menu, with Canon, Sony, Nikon and Casio camera presets.
Native Metal camera simulation includes sensor noise, lens softness, CCD bloom, sharpening and JPEG artifacts.
Fixed 16:9 preview supports playback, touch comparison and dragging the picture. Apply renders the effect and current audio mix into the clip; Undo restores the source.
Installation downloads now use pCloud, with a compact installation page and persistent release history.
Build 172 · 1.0
Centered the Original label at the top of fullscreen when holding to compare.