Speech to text: transcribe audio and video
Turn the speech in MP3, WAV, M4A, MP4, MOV and other files into SRT or VTT subtitles or plain text with Whisper, an AI speech recognition model that runs in your browser. It detects the language, or translates the speech into English.
Process locally
Files stay in this browser session. Closing or clearing releases them.
Choose files
MP3, WAV, M4A, AAC, OGG, FLAC, WebM, MP4, MOV, MKV, up to 5
Drop files here, or
Adjust settings
Optional: Spoken language, Text in, File format and more
Run
Processes in this browser; cancel at any time
About this tool
Turn the speech in audio and video files into subtitles or text: MP3, WAV, M4A, AAC, OGG, FLAC, WebM, MP4, MOV and MKV files become SRT or WebVTT subtitles or plain text, for example to subtitle a video, write up an interview or a meeting, or get notes from a lecture. The language is detected automatically, and speech in any language can also be translated into English.
Recognition is done by Whisper base, OpenAI’s open multilingual speech recognition model (74 million parameters, MIT licence), running in your browser, so your files are never uploaded. The model, about 79 MB, downloads the first time and is then kept by the browser. On a fast desktop it transcribes about 3 times faster than real time, on a laptop roughly in real time. Up to 5 files, 80 MB in total and at most 60 minutes each, go in one run.
How to use it
- Choose or drop up to 5 audio or video files: MP3, WAV, M4A, AAC, OGG, FLAC, WebM, MP4, MOV or MKV, 80 MB in total.
- Keep “Spoken language” on “Detect automatically”, or pick the language if you know it.
- Under “Text in”, keep “The language spoken”, or choose “English (translated)” to get English text from speech in any language.
- Pick the “File format”: “SRT subtitles” (the default), “WebVTT subtitles” or “Plain text”. For Chinese speech, choose “Simplified” or “Traditional” as the “Chinese script”.
- Select Run; the first run downloads the model. Read all the text in the preview of transcript.txt, then download the files one by one or all together as a ZIP.
Formats and limits
- Input
- MP3, WAV, M4A, AAC, OGG, FLAC, WebM, MP4, MOV, MKV, up to 5 files
- Output
- X-SUBRIP, VTT, TXT, Plain text
- Size limit
- 80 MB per file
- Settings
- Spoken language; Text in (The language spoken, English); File format (SRT subtitles, WebVTT subtitles, Plain text); Chinese script (Simplified, Traditional)
- Privacy
- Files stay in the current browser session. No file bytes are uploaded.
- Price
- Free, no sign-up
Common questions
Are my recordings uploaded?
No. Only the model is downloaded from this site; the sound is decoded and recognized in your browser and stays on your device.
How accurate is it?
Whisper base is a small model. It is very good on clear English and usable on clear speech in other major languages, but makes more mistakes with accents, background noise, several people talking at once, music and less common languages, and names and technical terms are often misspelt; Chinese gets more errors than English. Clean a noisy recording first with “Remove background noise from audio”, and check the text before you publish it.
Which languages can it transcribe and translate?
Whisper knows 99 languages, and “Detect automatically” covers them all; the “Spoken language” list names 19 of them. The language is detected from the first 30 seconds and kept for the whole file, so if a recording starts with music or with another language, pick the language yourself. Translation goes into English only.
How long can a recording be, and how long does it take?
Each file may last up to 60 minutes; cut longer ones with “Trim audio” or “Trim video”. All files together may be 80 MB, so for a large video, take the sound out first with “Extract audio from video”. A run may take at most 30 minutes, not counting the first model download: on a fast desktop that is about 90 minutes of recording, on a laptop that works in real time about 30.
Why is the Chinese text in Simplified characters?
Whisper writes Chinese in Traditional characters. With “Simplified”, the default, the characters are converted with OpenCC, one by one, so the words stay as they were spoken; “Traditional” keeps Whisper’s characters. Either way, commas, question marks and similar marks next to Chinese characters become Chinese punctuation.
What files do I get, and how are the subtitles timed?
Each recording gives a subtitle or text file with its name, such as interview.srt. Each subtitle is a segment of about 2 to 10 seconds, timed to 0.02 s as Whisper predicts it; the recording is processed 30 seconds at a time, and stretches of silence or music without speech are skipped. transcript.txt holds all the text without timestamps, headed by each file’s name when there are several, and goes on to the next step of a workflow, for example a translation.
Related tools
Extract audio from video (MP4 to MP3)
Take the sound out of MP4, MOV, MKV or WebM video and convert it to MP3, M4A, OGG, WebM or WAV, or keep the track unchanged in its own container, optionally trimmed. Files are not uploaded.
Remove audio from video
Mute MP4, MOV, WebM and MKV videos by removing the audio track. The picture is copied without re-encoding, so its quality, resolution and rotation stay exactly the same.
Convert audio to MP3, M4A, OGG or WAV
Convert MP3, WAV, M4A, AAC, OGG, FLAC and WebM audio to MP3, M4A, OGG, WebM or WAV, with bitrate, sample rate and channels. MP3 is encoded with LAME and the other formats by the browser; files are not uploaded.
Remove background noise from audio
Clean up voice recordings, podcasts, lectures and calls by removing hum, hiss, fans, traffic and other background noise with RNNoise, a small neural network that runs in your browser. Save as MP3, M4A, OGG, WebM or WAV.
Join audio files
Merge MP3, M4A, WAV, OGG, FLAC and other audio files into one track, in the order you add them, end to end, with a pause between them or with a cross-fade. Save the result as MP3, M4A, OGG, WebM or WAV.
Voice recorder
Record your voice or any sound from a microphone, with a level meter, pause and resume, then save it as MP3, WAV or the browser’s own WebM or M4A.