Audio to text

Convert audio to text, accurately

Upload a recording and read it a few minutes later: punctuated, split into paragraphs, with a timestamp on every line. Works in 50+ languages, with files up to 10 hours.

The short answerTo convert audio to text, upload the file to SpeakToWords, choose the language (or leave it on automatic) and wait a few minutes - a one-hour recording is usually ready in under six. You get a punctuated transcript you can edit while listening and download as TXT, DOCX, SRT or VTT.

Transcribe a file free See plans 15 free minutes, no card. 50+ languages. Files up to 500 MB.

How it works

Upload the recordingDrag in any audio file - MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, WMA, AIFF, AMR, 3GP, CAF, WebM, MP4 and MOV, up to 500 MB.
Let it transcribeA current speech model writes it out with punctuation and timestamps; you get an e-mail when it is done.
Check and exportClick any line to hear it, fix a word, then download in the format you need.

What makes the transcript usable

Raw speech recognition gives you a stream of words. What makes it worth reading is what happens around it: punctuation, paragraphs, and a way to check a line against the recording without hunting through the file. Every line here carries the second it was said; click it and the audio plays from there.

Which audio works best

Clear speech close to the microphone beats any expensive gear in a noisy room. One or two people taking turns transcribe almost perfectly; a crowded table, music under the voice or a phone call on speaker cost accuracy. If the recording matters, record a little closer to the speaker and away from air conditioning and clattering cups.

Pricing

A free account transcribes 15 minutes with no card and shows the opening of each transcript; the full text unlocks with the 7-day trial or a plan from €9.99 a month for 500 minutes (€4.00 for 3 months until 6 December). Pro is €14.99 a month for 1,500 minutes (€6.00 for 3 months until 6 December), for long recordings, AI summaries and translation.

Questions

Is it free?

How accurate is it?

On clear speech, upwards of 99% of words come back right, punctuated and with a timestamp on every line. Background music, people talking over each other, strong accents and unusual names cost accuracy; the editor plays the audio from any line so you can fix those quickly.

Does it label who is speaking?

Not yet. Every line carries the second it was said, so you can jump to it and listen, and add names in the editor. If speaker labels are essential, see the comparison pages for tools that offer them.

What happens to my file?

It is processed for you only, encrypted in transit, and never used to train anything.

Try it on your own recording

Start free