Transcribe audio without uploading
Transcribe audio with nothing uploaded. The speech recognition runs on your own machine, so your recording never leaves your device.
Advanced options
Your subtitles
Click any line to jump the player. Edit text directly — changes apply to the download.
Why "without uploading" matters
Virtually every free transcription service works the same way: you upload the audio, their server processes it, you download the text. It is convenient, and it means your recording passes through infrastructure you do not control.
Running the model locally removes that step entirely. The trade-off is speed — a browser is slower than a data centre GPU — but for anything sensitive, or anything you would rather not hand over, that trade is usually worth it.
You can verify this yourself: open the Network tab in your browser's developer tools before starting. You will see the one-time model download and nothing carrying your audio.
FAQ
How can I be sure nothing is uploaded?
Open DevTools, go to the Network tab and run the tool. The only outbound request is the model download on first use. After it is cached you can disconnect entirely and the tool keeps working.
Is it slower than cloud services?
Yes, roughly 0.6× realtime on a desktop — a 2 minute recording takes about 3 minutes. There is no queue, though, so short files often finish sooner overall.
Do you keep a copy of my transcript?
No. There is no account and no server-side storage; the transcript exists only in your browser until you download it.
Does it work offline?
After the first model download, yes — the model is cached by your browser.