🎙️ Offline
Transcriber
✓ Audio & video → timestamped transcript
← bunka.be
📖 Manual
Model
Tiny — fastest (~75 MB)
Base — balanced (~145 MB)
Small — most accurate (~480 MB)
Language
Auto-detect
English
Dutch (Nederlands)
German (Deutsch)
French (Français)
Task
Transcribe (same language)
Translate → English
Acceleration
Auto (WebGPU if available)
Force WebGPU
CPU (WASM)
Drop an audio or video file here
, or click to choose.
Nothing is uploaded — the model runs locally. First run downloads the model once.
Transcribe
Clear
Working…
Segments
Plain text
⬇ SRT
⬇ VTT
⬇ TXT
⬇ JSON
⧉ Copy text