VOICE.TRANSCRIPT --:--:--
speech to text · mp3 to text · live mic dictation · whisper in your browser · webgpu · up to large-v3-turbo · srt + vtt export · multilingual · transformers.js · speech to text ·
← archive · on-device whisper · srt + vtt
VOICE / TRANSCRIPT
speak into the mic or drop an mp3 get text
audio files are transcribed on-device — nothing leaves your browser. the live mic uses the browser's own speech engine, which in chrome runs on google's servers.
// 01
LIVE MIC
web speech api
ready
transcript
0 words
// 02
AUDIO FILE
whisper · on-device
transcript
0 words
live mic · browser SpeechRecognition (chrome · edge · safari — chrome routes mic audio through google)
audio file · whisper · tiny → large-v3-turbo (pick a model in the panel · fully on-device)
runtime · transformers.js v3 · onnx · wasm (runs in a worker — the page stays alive)
polish · gemma 2 2b · huggingface inference api (rate-limited)
first use of a model downloads it, then it's cached. large-v3-turbo needs webgpu (chrome / edge) — it's the superwhisper-grade one.