Tweezing — Precision Word-Level Video and Audio Editor
Search spoken words instantly. Cut exactly at syllable boundaries. Turn speech into short videos in seconds.
Millisecond Precision
Whisper-based word alignment calculates exact phoneme boundaries; clean cuts at exact word boundaries without clipping consonants.
Local DuckDB Storage
All speech transcripts, pitch contours, and indexes are stored in local DuckDB. Search thousands of hours of audio instantly without cloud dependencies.
100% On-Device & Private
Your raw video and audio files never leave your computer. Transcription, editing, and rendering run entirely on your CPU and GPU.
Engineering & Transparency
Tweezing combines an Electron desktop interface with a high-performance Python engine (FastAPI, DuckDB, Whisper) over local IPC. No fabricated claims, no exaggerated vanity metrics.
Desktop Shell: Electron (pure ES modules, 0 bundled bloat)
Engine: Python 3.11 sidecar (FastAPI + PyInstaller frozen executable)
Database: DuckDB (vector similarity & full-text search)
Speech Alignment: OpenAI Whisper (on-device inference)