Elide — AI video editor, open source
Manually cutting filler words, awkward pauses, false starts, and dead air out of videos is tedious. Creators spend hours on what should be automated — and cloud tools mean uploading private footage.
Local AI pipeline
FFmpeg extracts audio → embedded whisper-rs transcribes with word-level timestamps → a local LLM (via Ollama) decides which segments to cut → FFmpeg reassembles the clean video.
Why Rust
A 30-minute video means heavy transcription, inference, and encoding. Rust gives C-level speed with memory safety — and the entire pipeline ships as a single binary. No Python env, no Docker, no servers.
Why local
Video is personal and often proprietary. Everything runs on the user's machine: zero ongoing cost, zero data leaving the device.
the hard part Core pipeline works end-to-end with CI/CD (cargo fmt, build, clippy). Now refining the LLM prompt for better edit decisions and handling edge cases: overlapping speech, music segments, intentional pauses.
- Rust
- whisper-rs
- LLM Inference
- FFmpeg
- Ollama
- CI/CD