Audio Transcription
Turn recordings into text with OpenAI's Whisper model — running entirely in your browser. Private, free, no per-minute fees.
Private speech-to-text
Meeting notes, interview quotes, voice memos — transcribing them with cloud services means uploading audio that may contain sensitive conversations. This tool downloads OpenAI's open Whisper model to your browser once and transcribes on your own device: the audio is never sent anywhere, there are no minutes limits, and no per-minute fees.
Frequently asked questions
Which model should I pick?
Tiny is the fastest and works well for clear English speech. Base is more accurate on noisy audio and supports more languages at similar speed on modern devices. Larger variants are more accurate but noticeably slower in a browser.
Why does the first transcription take a while to start?
The model (~40–150 MB depending on selection) downloads to your browser once and is cached. Later transcriptions start immediately.
Languages?
The multilingual models transcribe 90+ languages, and Tiny/Base English variants are tuned for English. Output language follows the audio's language automatically.