Media Transcriber
Turn a video or audio file into text. Transcription runs 100% locally in your browser (Whisper via transformers.js) — your file is never uploaded to any server.
1. Choose a source
Fetching a URL only works if that server allows cross-origin requests (CORS). If it fails, download the file yourself and use "Upload a file" instead — see the README for why.
2. Options
The transcript is always kept in the source language (never auto-translated to English). Auto-detect works from the first ~30 seconds of audio and can occasionally guess wrong — especially on short clips, clips that open with music/silence, or heavily accented speech. If the output comes back in the wrong language, re-run with the language typed in explicitly. For names, numbers, or official/legal text where every character matters, use the Small model and proofread the result — no speech-to-text model, this one included, is 100% reliable on proper nouns.
3. Progress
Idle.