What AI model powers the transcription?

This tool uses OpenAI's Whisper models — Whisper Tiny for Fast mode and Whisper Large v3 Turbo for Accurate mode. Both run entirely in your browser via @huggingface/transformers (transformers.js) using WebAssembly. Your audio never leaves your device and is never uploaded to any server.

Why is the first transcription slow?

The Whisper model must download on first use — about 75 MB for Fast mode and roughly 400 MB (quantized to int4) for Accurate mode. After the initial download the model is permanently cached in your browser, so all subsequent transcriptions start instantly without any additional download.

Which languages are supported?

Fast mode (Whisper Tiny) is optimised for English and works reasonably well for other common languages. Accurate mode (Whisper Large v3 Turbo) supports over 100 languages including Spanish, French, German, Arabic, Chinese, Japanese, Hindi, Urdu, Portuguese, Russian, and many more. The language is detected automatically from the audio content.

Free Audio Transcription Online — AI Powered

Transcribe any audio file to text using Whisper AI, running entirely in your browser. Supports MP3, MP4, WAV, M4A, and WEBM. Free, no signup, no server upload.

How it works

1Upload any audio or video file by dragging it in or clicking to browse. Supports MP3, MP4, WAV, M4A, and WEBM up to 200 MB.
2Choose your quality mode: Fast (Whisper Tiny, ~75 MB) for quick English transcription, or Accurate (Whisper Large v3 Turbo, ~400 MB) for best results across 100+ languages. The model downloads once on first use and is permanently cached in your browser — no re-download on future visits.
3Click "Transcribe Audio" — Whisper AI runs locally on your device and converts the speech to text. The detected language appears below the transcript. Copy or download the result as a .txt file with a single click.