How to transcribe audio to text for free
2026-07-19
Transcription used to mean either typing it out yourself or paying per minute. Neither is necessary for most recordings any more. Here is the short version: upload the file, wait a couple of minutes, copy the text.
The quickest route
- Open the home page and drop your file on it. It starts uploading immediately and a window shows the progress, then the file’s name, size and length.
- In that same window, leave the language on Detect automatically unless the recording is noisy, in which case naming the language helps.
- Keep Label speakers on if more than one person is talking, and say how many. If you are not sure, press I don’t know — detect it instead.
- Press Transcribe and read it as it appears.
- Put names to the voices with Name speakers, then copy the
transcript or download
.txtor.vtt.
Free, up to 100 MB and 60 minutes per file, no signup.
What you can upload
Most audio and video formats: mp3, m4a, wav, ogg, opus, flac, aac, mp4, mov, mkv, webm. Video files are fine — only the audio track is read, so there is no need to extract it first.
How to get a better transcript
Transcription quality is mostly decided before you upload anything. In rough order of impact:
- Record close to the speaker. A microphone two feet away beats any amount of post-processing.
- Avoid overlapping speech. Crosstalk is the single hardest case, both for the words and for the speaker labels.
- Name the language if the recording is short or noisy. Auto-detection works from the audio itself and has less to go on in both cases.
- Do not pre-compress. Uploading an already heavily compressed file throws away detail the transcriber could have used.
Why speaker labels matter
For a single voice — a lecture, a voice memo — plain text is fine. For an interview or a meeting it is close to unusable: you can read every word and still not know who committed to what. Speaker labels turn a transcript into a record.
We do this with diarization: the audio is split into speech segments, each one gets a voice fingerprint, and segments are clustered by voice. It works well on clear recordings with distinct voices, and struggles with very short interjections, where there is not enough audio to fingerprint.
Clustering can only tell voices apart, never name them, so they arrive as
Speaker 1, Speaker 2 and so on. Once the transcript
is done, Name speakers lets you replace those with real names —
each voice is listed with how much it talks and its first line, which is usually enough
to work out who is who. The names go into the .vtt and .txt
you download, not just the page.
What happens to your file
It is deleted as soon as the transcript is ready. The transcript is deleted after 24 hours. There is no account, so nothing links either to you.