How to use Audio Transcription and Captions
- Choose an English-speech PCM16 WAV, no longer than 30 seconds. Complete browser verification, then select Transcribe English speech to upload and process it.
- Keep the original open in your audio player. Correct each caption and its start/end seconds against what you hear; use Previous segments and Next segments when the editor has more than 25 cues.
- Add or remove segments when needed. Times must stay ordered without overlap, and each caption must contain one plain-text line.
- Select Export current result to save transcript.txt, captions.srt and captions.vtt. Check the cues with the recording in your target player.
Example: Audio Transcription and Captions
Review captions for a short spoken instruction before exporting them.
Supported inputs and limits
- Server-side pinned tiny.en English Whisper model.
- One PCM16 WAV, mono/stereo, 8–48 kHz, up to 20 MiB and thirty seconds.
- At most 300 ordered, non-overlapping caption segments with one plain-text line each.
- No translation, speaker identification or word-level alignment.
- Noise, accents and silence can cause invented or missing words.
More details
Jobs expire after fifteen minutes.
Where your input is processed
Your file is uploaded to an isolated server worker after Cloudflare human verification. The verification request does not include the file. Inputs are removed after processing. Downloads belong to this browser session and expire after 15 minutes. The worker removes expired jobs automatically; if it stops, temporary files can remain until cleanup resumes or the scheduled fallback runs.
A readable transcript still needs timing review
Check names, negatives, numbers and pauses against the recording before fixing punctuation. A plausible sentence may contain words that were never spoken, particularly over noise or silence. Place each cue where its speech begins and ends, then test the exported captions in the player you intend to use. The editor supports up to 300 segments and one plain-text line of up to 2,000 characters per segment. It does not identify speakers or align individual words.
Questions about Audio Transcription and Captions
Can I correct the result?
Yes. Edit every text segment and its start/end seconds before exporting.
Which languages work?
This initial tiny.en model supports English only.
Is the transcript reliable evidence?
No. It can omit or invent words; check the recording and every cue.