Transcribe audio to text
Turn recordings into a searchable PDF transcript.
Transcribes the speech in an audio or video file into a searchable PDF, in your browser.
How to use it
- Choose an audio or video file: MP3, WAV, M4A, AAC, OGG, Opus, FLAC, WebM, MP4 or MOV.
- Choose Standard (faster) or Accurate (slower), and the spoken language, or let superPDF detect it.
- Select Transcribe recording. The first time, superPDF shows the model’s size and downloads it only after you select Download; for Standard you can choose the faster version (209 MB) or the smaller one (80 MB). The transcript appears as it is written; select Stop to keep what is done.
Options
| Option | Values | Default |
|---|---|---|
| Files | Accepts MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, WEBM, MP4, MOV | |
| Transcription | One of Standard (faster), Accurate (slower) | Standard (faster) |
| Spoken language | One of Detect automatically, Arabic, Chinese, Czech, Dutch, English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Swedish, Turkish, Ukrainian, Vietnamese | Detect automatically |
Output
A PDF transcript with a timestamp heading every minute or so, a text file, and SRT and WebVTT subtitle files for video players.
How it works
superPDF cuts the recording into short pieces at the quietest moments, so words are rarely split, and transcribes each piece with timestamps. To detect the language, it listens to the first 30 seconds of speech. The PDF uses timestamps as headings, so you can search it and jump to a moment.
Limits
No file size or page limit. Everything runs on your device, so very large files are fastest on a computer. On a fast computer, Standard transcribes five minutes of clear speech in about 15 seconds with the faster version and 50 seconds with the smaller one; Accurate takes about twice as long as the smaller Standard version. Slower computers take longer. Accuracy drops with background noise, overlapping speakers, accents and specialist words, and the transcript doesn’t say who is speaking. It can occasionally repeat or invent a phrase in long silences, so check names, numbers and quotes against the recording. The PDF uses standard fonts, which can’t show non-Latin scripts: those characters appear as “?”, and the text, SRT and WebVTT files have the full transcript.
Privacy
Runs in your browser. Your files are processed on your device and never uploaded.