← all repositories
MartinDelophy/ai-video-editor

A CapCut clone where the AI is your co-editor, not a black box

Timeline Studio runs a full multi-track video editor in the browser with local WebGPU AI — and lets agents like Codex edit the same timeline you do.

★1k stars JavaScript Creative · Design
ai-video-editor
Too new to rank. Velocity needs a few days of history.
star history

What it does

Timeline Studio is a local-first, browser-based video editor with a CapCut-style multi-track timeline. The twist: AI agents can edit the same real timeline as you, via a WebMCP interface — a demo shows Codex proposing a beat-template edit that a human then reviews, tweaks, and exports. On top of that sit a pile of in-browser AI features: multilingual TTS voiceovers, music generation, automatic captions, watermark removal, 4× video restoration, vocal separation, and talking-avatar generation.

The interesting bit

Everything heavy runs locally — Whisper, Stable Audio, MI-GAN, LivePortrait and friends are ONNX/WebGPU models lazy-loaded and cached by a service worker, with no project media uploaded to a backend. Model delivery is dual-mirrored on Hugging Face and ModelScope with shared cache identities, so a fallback switch doesn’t re-download the same revision. That’s unusually thoughtful plumbing for a browser app.

Key highlights

  • Agents and humans share one editable timeline through WebMCP; there’s a documented “Timeline Studio Skill” workflow
  • Local AI stack: Hojo/Kokoro/Piper TTS, Stable Audio 3 music, Whisper captions, YOLOS+MODNet smart framing, MI-GAN repair, NanoVSR restoration, JoyVASA+LivePortrait avatars
  • Deterministic offline export to MP4/MOV, including 4K with on-demand media loading to limit memory pressure
  • Editable beat templates (up to 48 replaceable slots), keyframeable color grading, 120 searchable sound effects
  • Runs as a PWA; an “Anna” cloud edition adds project continuity and plan-based AI edits

Caveats

  • The Anna edition’s planning works from clip metadata only — no frame or audio analysis — and its AI features depend on the container’s capabilities
  • Deep-synthesis features (talking avatars) carry explicit legal-use restrictions; the project is framed as research/learning tooling
  • Voice models ship with mixed upstream licenses — check the model card before commercial use

Verdict

Worth a look if you’re curious where agent-assisted creative tools are heading, or you want a serious video editor that never uploads your footage. If you need broadcast-grade NLE features, this is still a browser app with browser-app limits.

Frequently asked

What is MartinDelophy/ai-video-editor?
Timeline Studio runs a full multi-track video editor in the browser with local WebGPU AI — and lets agents like Codex edit the same timeline you do.
Is ai-video-editor open source?
Yes — MartinDelophy/ai-video-editor is open source, released under the MIT license.
What language is ai-video-editor written in?
MartinDelophy/ai-video-editor is primarily written in JavaScript.
How popular is ai-video-editor?
MartinDelophy/ai-video-editor has 1k stars on GitHub.
Where can I find ai-video-editor?
MartinDelophy/ai-video-editor is on GitHub at https://github.com/MartinDelophy/ai-video-editor.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.