Pluggable transcription for mcut: provider interface, normalized transcripts, SRT/VTT, caption generation

Programmable editor operators for mcut: user-level actions over the timeline engine

mcut as an MCP server: every editor command and operator becomes a tool any MCP client can call

mcut transcription provider backed by the Vercel AI SDK — use any AI SDK transcription model (OpenAI, Deepgram, Groq, ElevenLabs, ...)

Pure canvas2d frame renderer for mcut projects — one compositor for interactive preview and deterministic export

On-device mcut transcription provider: Whisper via Transformers.js in a worker — WebGPU-gated, chunked streaming, repetition/VAD guards

React bindings for mcut: EditorProvider, hooks, PlayerCanvas with selection overlay, playback clock, audio preview

mcut on the command line: scaffold, validate, summarize, and batch-edit project documents headlessly

mcut transcription provider backed by AssemblyAI: word timestamps, confidence, and speaker labels

Media I/O for mcut on Mediabunny: probing, thumbnails, audio extraction, frame sources, client-side export with a pluggable container-format registry (MP4/WebM/MKV built in)

Headless video timeline engine: project model, serializable commands, undo/redo, reactive stores