Audio in. Text out.
Nothing leaves your Mac.
Drop in a recording and get an accurate Polish or English transcript in minutes, running entirely on your own machine.
This demo is simulated in the browser with sample clips. The real app transcribes your own files locally.
Features
Fully offline
whisper.cpp runs on your own machine. Your audio never leaves it, and there are no accounts or API keys.
Polish & English
Built on the multilingual large-v3-turbo model. Pick a language or let it auto-detect.
Fast on Apple Silicon
Runs on the GPU via Metal. About 8× real-time on an M4: an hour of audio in roughly eight minutes.
Live progress
A real progress bar and ETA, driven by the transcriber's own output, not a spinner.
Fewer hallucinations
Voice-activity detection skips silence, and context is reset per window to curb repetition loops.
Any audio format
wav, mp3, m4a and most others. ffmpeg handles conversion. Export as Markdown or copy.
How it works
- 01
Drop a file
Drag in any audio recording and choose a language.
- 02
Transcribe locally
ffmpeg converts it, whisper.cpp transcribes it on your GPU.
- 03
Copy or download
Get clean text, ready to paste or save as .md.