Show HN: I built an open source video editor that you can control with an LLM
Donkey Cut is a free, open-source video editor that runs in the browser, positioning itself as a CapCut alternative with AI-powered editing capabilities The platform supports both cloud-hosted and local projects, giving users full control over where their media files are stored and processed An integrated AI assistant can generate video clips, images, voiceovers, and music directly within the editor, with iterative chat-based refinement A macOS companion app enables local processing via a bundle
Analysis
TL;DR
- Donkey Cut is a free, open-source video editor that runs in the browser, positioning itself as a CapCut alternative with AI-powered editing capabilities
- The platform supports both cloud-hosted and local projects, giving users full control over where their media files are stored and processed
- An integrated AI assistant can generate video clips, images, voiceovers, and music directly within the editor, with iterative chat-based refinement
- A macOS companion app enables local processing via a bundled Cut engine, on-device speech transcription, screen recording, and integration with existing Claude/Codex logins
- The architecture separates cloud projects (hosted APIs, Postgres, R2 storage) from local projects (browser storage or ~/Movies on Mac), with the editor UI remaining consistent across both modes
Why It Matters
Donkey Cut represents a growing trend of open-source, AI-native creative tools that lower the barrier to professional video production by combining traditional editing workflows with generative AI capabilities. For AI practitioners and developers, it demonstrates a practical hybrid architecture where cloud and local processing coexist, offering a blueprint for privacy-conscious AI applications that don't force all computation into the cloud.
Technical Details
- Hybrid Architecture: Projects can be cloud-based (hosted APIs, Supabase Postgres via Prisma, R2 storage) or local (browser storage or Mac disk at ~/Movies), with both types visible on a unified home screen
- macOS Companion App: Built with Swift, runs a local Cut engine on 127.0.0.1, bundles ffmpeg for processing, performs on-device speech transcription, and connects to existing Claude/Codex authentication
- AI-Integrated Editing: The editor features a multi-track timeline with captions, music, and effects; the AI assistant generates media (video, images, voiceovers, music) that integrates seamlessly as timeline assets
- Iterative Generation Workflow: Users can describe shots in chat and refine outputs iteratively, as demonstrated by example projects like "The Railway Mystery" (Franco-Belgian comic style with generated shots and brass-and-strings score) and "City poster series" (animated hand-painted travel posters)
- Tech Stack: Next.js frontend, Supabase Postgres, Prisma ORM, Apache 2.0 license; local development requires npm, Supabase credentials in .env, and Swift for the macOS app
Industry Insight
- The hybrid cloud/local model addresses growing user demand for data privacy without sacrificing accessibility—this dual-mode approach could become a standard pattern for AI-powered creative tools, especially as regulations around data residency tighten
- Integrating generative AI directly into the editing workflow (rather than as a separate tool) represents a shift toward "AI-native" creative suites; competitors should consider embedding generation capabilities natively rather than relying on external API calls
- The open-source, freemium model (free editor, paid cloud storage and AI generation) provides a sustainable revenue framework while building community adoption—this pricing strategy could be replicated for other creative AI tools seeking to balance accessibility with monetization
Disclaimer: The above content is generated by AI and is for reference only.