Voice to text, privately
Voice messages, recordings, lectures, videos — get text with timecodes. Whisper AI runs right in your browser: the audio never leaves your device.
Settings
Stays on your deviceRecord from microphone0:00
Tip: in Telegram, long-press a voice message → Save to Files / Download, then drop it here.
Transcript
Click a line to play from there
0bytes of audio are uploaded. Whisper (OpenAI, open source) runs in this tab via WebGPU, or on the processor if your browser has no WebGPU.
1×the model is downloaded once (Fast ≈ 80–200 MB, Accurate ≈ 250–400 MB) and cached by the browser. Next time it starts instantly, even offline.
99languages are understood, auto-detected by default. Choose the language yourself if a short message comes out in the wrong one.