Multilingual AI voiceovers
Chinese voices use Piper/VITS browser ONNX. English voices use Kokoro 82M ONNX. Browser Piper voices are also available for German, Spanish, French, Italian and Brazilian Portuguese.
Product capabilities
A local-first AI video editor that runs in your browser
Timeline Studio combines AI voice generation, automatic captions, talking portraits, visual editing and a multi-track timeline in one installable web app. Core editing and inference workflows run on the user's device; model files are downloaded on demand and cached for repeat use.
AI creation
Automatic captions
Whisper small q8 runs in a WASM worker. Caption timing is adjusted toward nearby audio energy, with conservative cleanup for high-confidence Chinese homophone errors.
Talking portraits
JoyVASA drives facial motion from audio and LivePortrait renders the result with a WebGPU path, including 256px preview and 512px quality modes.
Browser vision
YOLOS tiny detects subjects and MODNet creates portrait masks for smart cropping, caption avoidance and background removal in preview and export.
Timeline and visual editing
Multi-track timeline
Arrange visuals, stickers, captions, source audio, generated voiceovers and music. Split, duplicate, move, resize, hide, lock and delete selectable clips.
Visual keyframes
Animate scale, horizontal and vertical position, rotation and opacity with grouped or property-level keyframes and independent linear interpolation.
Masks and effects
Use rectangle, rounded-rectangle and circular masks with canvas-accurate position, size, radius, feathering and inversion, plus filters and transitions.
Video speed and audio sync
Adjust clip-scoped playback speed and choose whether extracted source audio remains linked to the visual clip or can diverge independently.
Export and privacy
| Video output | MP4 when the browser supports the required recording or conversion path; WebM is the fallback. |
|---|---|
| Resolution | Export controls include 720p, 1080p, 1440p and 2160p choices, subject to device capability. |
| Frame rate | 30fps and 60fps options are available, subject to browser and device performance. |
| Data model | Media is selected from the user's device and processed in the browser. Third-party model hosting may receive normal network requests when model assets are downloaded. |
| Price | The public browser editor is free to open and the source code is released under the MIT License. |