Audio to Text
Drop a recording and get a written transcript, with punctuation and paragraphs. The speech model runs in your browser, so the recording never leaves your device.
Transcribed on your device by an AI speech model. Your file is never uploaded.
Transcribe in three steps
Drop your audio
MP3, WAV, M4A, AAC, FLAC or OGG, up to 30 minutes.
Pick the language
Or leave it on automatic detection.
Transcribe and copy
Copy the text or download it as .txt.
Good recordings make good transcripts
Whisper copes well with ordinary microphones, but the closer and clearer the voice, the better the text. Background music and overlapping speakers cause most mistakes.
Voice memos from iPhone are M4A files and work as they are. Paragraph breaks are added wherever the speaker pauses for two seconds or more.
How the transcription works
Transcription is done by Whisper, the open-source speech recognition model from OpenAI, running inside your browser. The model downloads once (40 to 250 MB depending on the accuracy you pick) and is cached, so later files start straight away. Your video is never uploaded.
With a recent Chrome or Edge on a computer with a graphics card, it runs on the GPU and takes a fraction of the video's length. Without one it runs on the CPU, which can take about as long as the video itself. Files up to 30 minutes work.
Whisper understands about 100 languages and adds punctuation. Accuracy is good on clear speech, and lower with music, heavy accents or several people talking at once: check names and numbers before you publish.
Frequently
Asked Questions
Can I transcribe an iPhone voice memo?
Yes. Voice memos are M4A files: drop one in.
Is my recording uploaded?
No. It's transcribed in your browser.
Does it tell speakers apart?
No. The transcript doesn't label who's speaking.
How long can the recording be?
Up to 30 minutes per file.