- Home
- AI Video Editing Software
- Dawn Cut
Dawn Cut
What it is
Dawn Cut is a free, open-source video editor for macOS that keeps projects local. It lets you edit video through transcript text, subtitles, and automatic silence removal. Deleting a range of words cuts the corresponding footage and closes the gap. Local whisper.cpp provides word-level transcription and Korean auto-subtitles. Editing tools include filler-word detection, subtitle review and correction, color grading, reframing, overlays, and style packs. The project lists SRT and chapter export, deterministic EDL export, and video rendering through an FFmpeg sidecar. An experimental headless MCP server can allow Claude Desktop, Cursor, and other external agents to operate on .dawn projects. Natural-language editing uses a local llama.cpp planner, requires a one-time model download, and is bounded by a 1.5B model. The project describes Dawn Cut as 100% local, with no cloud or account, and releases it under the MIT license. It is labeled v0.1, with a deterministic editing core ready to use.
For and against
For
- Free plan with no subscription, account, or watermark
- Word-range edits cut footage and close gaps
- Local word-level transcription with Korean subtitles
- Silence removal, filler detection, and subtitle correction
- SRT, chapter, and deterministic EDL export
Against
- macOS only
- Natural-language planner is experimental and limited to 1.5B model
- One-time model download required for natural-language editing
- MCP rendering does not burn subtitles or overlays
The verdict
Dawn Cut combines transcript editing, local transcription, and conventional video tools in a free macOS package. Its experimental AI and MCP features have stated limitations, while the deterministic editing core is ready to use.
Who it suits
Dawn Cut suits macOS editors who want transcript-based video work without cloud storage or an account. It also fits users needing Korean auto-subtitles or local project processing.
Plans and prices
Free
No subscription · no account · no watermark
Read 8 Oct 2026 · github.com
On the record
- Automatic captions
- Yesgithub.com · read 21 Sep 2026
- Editing features
- Features include silence removal, filler-word detection, subtitle review and correction, color grading, reframing, overlays, and style packs.github.com · read 8 Oct 2026
- Experimental AI
- Natural-language editing uses a local llama.cpp planner and requires a one-time model download; planning quality is bounded by its 1.5B model.github.com · read 8 Oct 2026
- Exports
- The README lists SRT and chapter export, deterministic EDL export, and video rendering through an FFmpeg sidecar.github.com · read 8 Oct 2026
- Integrations
- The experimental headless MCP server can let external agents such as Claude Desktop and Cursor operate on .dawn projects.github.com · read 8 Oct 2026
- License
- The project is released under the MIT license.github.com · read 8 Oct 2026
- MCP limitation
- The experimental MCP render does not yet burn in subtitles or overlays, and its npx-installable package is not published.github.com · read 8 Oct 2026
- Privacy
- The project describes Dawn Cut as 100% local, with no cloud and no account.github.com · read 8 Oct 2026
- Product
- Dawn Cut is an open-source, local AI video editor for editing video through text, subtitles, and automatic silence removal.github.com · read 8 Oct 2026
- Requirements
- The quickstart lists Node 20 or later and pnpm 10, plus cmake and FFmpeg for transcription setup.github.com · read 8 Oct 2026
- Silence removal
- Yesgithub.com · read 21 Sep 2026
- Status
- The README labels the product v0.1 and says the deterministic editing core is ready to use.github.com · read 8 Oct 2026
- Supported platforms
- macOSgithub.com · read 21 Sep 2026
- Text-based editing
- Yesgithub.com · read 21 Sep 2026
- Text editing
- Deleting a word range cuts the corresponding footage and closes the gap.github.com · read 8 Oct 2026
- Transcription
- The app uses local whisper.cpp for word-level transcription and supports Korean auto-subtitles.github.com · read 8 Oct 2026
