Subtitle Desk Browser-only tool

ENGLISH SPEECH → PLAIN TEXT

Video to text.

Choose a video or audio file

English WAV, MP3, MP4 or WebM. One file at a time. Browser codec and memory limits vary.

Choose a file to begin.

Audio and video stay on your device. Only the recognition model and runtime are downloaded; your media is never uploaded.

What to expect

This page gives you reading text, split into paragraphs. It does not create subtitle timing. The transcription runs locally in overlapping 30-second windows. There is no fixed duration or file-size cutoff in this tool; a browser may still run out of memory. For video, the browser reads the audio during normal-speed local playback before transcription starts.

English only. The pinned model's encoder and decoder weights total 40,843,851 bytes (38.95 MiB), measured from the model files, plus tokenizer and runtime downloads. On this Windows test machine with a cached model, this page took 53 seconds for a 3-minute synthetic English recording and 182 seconds for a 10-minute recording. The latter produced 24 paragraphs and flagged one interval with no recognized speech. Your device and recording may differ.

The model can mishear words or miss quiet passages. Review the result against your source. If you need timed SRT or VTT instead, use auto captions. For existing subtitle files, use VTT to SRT, SRT to VTT, Text to SRT, SRT to TXT, VTT to TXT, the subtitle converter, ASS to SRT, SUP to SRT, the CPS checker, or the subtitle editor.