Transcribe audio to text

Upload the file above and download the transcript — no account, no email, no credits. It takes AAC, AIFF, FLAC, M4A, MP3, OGG and WAV up to 80MB and 20 minutes, and exports plain text plus SRT and VTT with nothing paywalled.

Yes. No sign-up, no email, no credits, no card. Exports aren't paywalled either — TXT, SRT and VTT all download without an account. 20 minutes per file and 2 per hour are the only limits, and they exist to keep the queue moving rather than to sell you an upgrade.

No. There's no app, no extension and no login — you upload the file in your browser and the transcription runs on our GPU server. The transcript appears on the same page with the download buttons already on it.

AAC, AIFF, FLAC, M4A, MP3, OGG and WAV, up to 80MB and 20 minutes per file. For video, use Video to Text instead — it takes MP4 and MOV directly without extracting the audio first.

Usually under a minute. The transcription server spins down when idle, so the first run after a quiet period spends about a minute starting up — which means a 30-second voice memo and a 10-minute podcast often take roughly the same wall time. Once it's warm, most files come back in seconds.

It depends far more on your recording than on the tool. Clean speech with one speaker close to a microphone comes back near-perfect; a phone recording of a meeting from across a table won't. We don't publish an accuracy percentage because a single number across every language, accent and recording condition wouldn't mean anything — and nobody publishing one shows their methodology.

Whisper large-v3, the largest model in that family, running on a GPU. Naming it means you can check it. Most free tools run a smaller variant and don't say which.

No — it's detected automatically. Setting it yourself helps for clips under about thirty seconds, heavy accents, or audio that mixes two languages, since detection works from the opening seconds.

Yes — choose English output and it translates as it transcribes, in one pass, at no extra cost. English is the only translation target available.

TXT is the words with no timing, for reading and searching. SRT and VTT are timed caption formats — SRT for video editors and most upload forms, VTT for HTML5 video on the web. All three are free to export.

Your upload is deleted as soon as processing finishes — not on a timer, and whether the job succeeded or failed. The transcript is available for 1 hour, then removed. Nothing is stored on disk afterwards. There are no accounts, so nothing is linked to you.

Not in one pass. Split it into sections first — the Silence Splitter cuts at natural pauses, which gives cleaner boundaries than cutting at a fixed time — then transcribe each section.