English | 中文
Timeline Studio is a local-first AI video editor that runs in the browser. It combines a CapCut-style multi-track timeline with browser-side AI voiceovers, automatic captions, vision tools, talking-avatar generation, and deterministic offline export.
Open the editor · Watch the demo · Hugging Face Space
Multilingual voiceover: Chinese Piper/VITS ONNX voices, English Kokoro 82M, and browser Piper voices for German, Spanish, French, Italian, and Brazilian Portuguese.Automatic captions: Whisper small q8 ONNX with waveform-aware timing and conservative Chinese recognition cleanup.Smart framing: YOLOS tiny subject detection and MODNet portrait matting for smart crop, caption avoidance, and background removal across images and complete videos.AI vocal separation: isolate vocals and place the instrumental stem on the music track without leaving the browser workflow.Digital human: JoyVASA audio-to-motion and LivePortrait neural rendering with WebGPU, 256px preview and 512px quality paths.Local-first inference: large models are lazy-loaded, revision-pinned, and cached by the service worker; supported workflows run without up project media to an editing backend.
- Contiguous main Visuals track plus timed picture-in-picture overlays.
- Direct canvas selection, movement, proportional resize, rotation, masks, filters, effects, animation, speed, and explicit keyframes.
- Captions, stickers, voiceover, separated source audio, and music on independent timed tracks.
- CapCut-style snapping, alignment guides, clip menus, split/duplicate/delete, timeline zoom, undo/redo, and portable
.timeline
projects. - Native media playback for a responsive preview; export uses a separate deterministic offline rendering path.
- WebCodecs MP4/WebM composition with shared preview/export geometry, audio mixing, captions, overlays, effects, and MediaRecorder fallback.
- Installable PWA with a cached app shell and multilingual UI.
The repository includes edit-timeline-studio, a Codex-compatible Skill for planning, executing, and verifying editable video timelines.
It helps an agent:
- inspect media and preserve the user's editing brief;
- describe reversible edits with stable clip IDs and explicit timestamps;
- operate the hosted or local editor through the browser compatibility path;
- validate declarative edit plans with
skills/edit-timeline-studio/scripts/validate_edit_plan.mjs
; - verify track placement, transitions, captions, overlays, audible audio, and final export artifacts;
- keep the editable
.timeline
project as the source of truth instead of returning only an opaque render.
The current Skill is honest about its boundary: browser-driven editing is available today, while the versioned headless command runner described in its command contract is the next automation layer.
Requirements: Node.js 20+ and a modern Chromium browser. WebGPU is recommended for the heaviest AI workflows.
git clone https://github.com/MartinDelophy/ai-video-editor.git
cd ai-video-editor
npm install
npm run dev
Open the local URL printed by Vite. The first AI run may download model files; later runs reuse the browser cache.
npm test
npm run build
npm run preview
Run the complete repository check with:
npm run check
The included netlify.toml builds with
npm run build
, publishes dist
, enables the cross-origin isolation headers required by browser AI/media workers, and provides the SPA fallback.
npx netlify-cli deploy --prod --dir=dist