Audio to Text & Subtitles

Turn audio or video into editable text and timed subtitles with Whisper — right in your browser. Your files never leave your device.

Source file

Drop an audio or video file here MP3, WAV, M4A, FLAC, OGG, MP4, WebM · Up to 250 MB / 20 min

On first use, your browser downloads the Whisper Tiny model (about 41 MB for CPU mode). Your audio stays on this device.

Transcript

Your timestamped transcript will appear here. Select a timestamp to jump to that point in the file.

Files and results stay on this page only. Clear the file or close the page to remove them. The Whisper Tiny model is downloaded once from Hugging Face and cached by your browser.

How to use Audio to Text & Subtitles

  1. Drop an audio or video file onto the page, or click “Choose file”.
  2. Pick the spoken language (or leave Auto-detect on), choose a processing mode, then click “Generate subtitles”.
  3. Click a timestamp to check a line against the audio, edit the text, then copy it or export SRT, VTT or TXT.

About this tool

Audio to Text & Subtitles runs Whisper Tiny — a small, open-source speech-recognition model — entirely in your browser, using Transformers.js and ONNX Runtime Web. On first use, your browser downloads the model (about 41 MB for CPU mode) from Hugging Face and caches it; after that, transcription itself runs on your device. Your audio and video files are never uploaded, and nothing you transcribe is sent anywhere.

Frequently asked questions

Are my files uploaded anywhere?

No. Transcription runs entirely in your browser and your audio or video file never leaves your device. The only download is the Whisper Tiny model itself: on first use your browser fetches it from Hugging Face (about 41 MB for CPU mode) and caches it for later. If Hugging Face is unreachable, the first transcription will fail — but your file still never leaves your device.

Which files are supported, and what are the limits?

Audio files in MP3, WAV, M4A, FLAC or OGG format, and videos with an audio track in MP4 or WebM format. One file can be up to 250 MB and 20 minutes long. If your browser can’t decode a file’s audio track, export it to WAV or MP3 first and try again.

Which spoken languages are supported?

Auto-detect, plus English, Chinese, Japanese, Korean, Spanish, French, German and Portuguese. Pick the language you hear in the recording — a wrong choice can make the transcript come out in the wrong language.

What’s the difference between WebGPU and CPU mode?

Auto mode tries WebGPU first and falls back to CPU if WebGPU can’t start or fails. WebGPU uses the full-precision (fp16 or fp32) model; CPU mode uses the smaller quantized model (about 41 MB). If your browser or device has no WebGPU — for example in a sandbox or on older hardware — CPU mode still works, just more slowly.

How accurate is it, and what can I export?

Whisper Tiny is a small, fast model: everyday speech usually comes out well, but it can mishear names, numbers and uncommon words, so proofread before publishing. Click any timestamp to jump to that point in the audio, edit each line in place, then copy the text or export SRT, VTT or TXT.

Like this tool?

Get Very Useful Tools Pro — one $6/month subscription for all tools: ad-free once ads launch, plus bigger batches and higher limits.

Go Pro — from $6