tscaps

What it is

tscaps is an open-source, client-side video editor built around captions. It runs entirely in the browser: drop a video in, transcribe it with in-browser Whisper, style the captions through CSS templates, and export the result frame by frame as a new video — with no upload, no account, and no backend.

How it works

Captions are HTML elements styled with CSS. In the editor, the same DOM is layered live over the video during playback. On export, the engine rasterizes that DOM at each frame's timestamp and composites it with the video frame. The browser is the renderer in both cases, which keeps the styling surface familiar and the export fast.

Key features

• In-browser transcription powered by Whisper, so audio never leaves the device
• CSS-based styling with a templated gallery of visual looks (drop a folder in to add a new style)
• Frame-by-frame export that bakes the styled captions into the final video
• Drop-in engine (@tscaps/engine on npm) for embedding the same pipeline in your own product without bringing the editor UI
• Project status: pre-1.0; pin to an exact version in production

Self-hosting

A pre-built Docker image is published as ghcr.io/francozanardi/tscaps-web:latest, which serves the static production bundle on nginx. You can also build from source via the included Dockerfile, or run the dev server directly with pnpm --filter ./apps/studio dev.

How it compares to Submagic

Both target short-form creators who want fast, AI-assisted captioning. Submagic is a hosted, paid service with extra features like auto B-roll, sound effects, and viral hooks. tscaps is the self-hostable, free alternative for users who only need the captions workflow — full ownership of the video, no watermark, no upload, and a styling model that is just CSS.

License

The web app is released under AGPL-3.0, and the standalone engine and templates under MIT, so the engine can be embedded in proprietary products while the UI stays copyleft.