Language
Transcriber
Any audio, any video, any language, transcribed locally.
A local transcription tool built on faster-whisper. FFmpeg normalises every input to 16 kHz mono before the model sees it, so codec and container stop mattering, and voice-activity filtering removes the text Whisper otherwise hallucinates over silence.
A backend ladder means a small GPU degrades to a slower path instead of crashing, which is the difference between a tool that works on the hardware you have and one that does not.
How it is built
- FFmpeg normalisation up front, so any container or codec is accepted.
- Voice-activity filtering on by default to cut hallucinated text over silence.
- A backend ladder so a 6 GB GPU falls back rather than failing.
- Writes text, SRT and JSON from a single run.
- Fully offline after the first model download.