How to convert audio to text
Audio to Text turns speech into a written transcript you can read, fix and reuse. It works with sound files and videos up to an hour long: podcasts, interviews, lectures, meetings, voice memos and talking-head videos. The speech recognition runs on your own device, in your browser, so the recording is never uploaded to a transcription service. The transcript appears as readable paragraphs with timestamps, and you can download it as plain text, a Word document or a subtitle file.
- 1
Add your recording
Drop in a sound file such as MP3, M4A or WAV, or a video such as MP4, MOV or WebM. Only the sound is used, so a video file is fine. Recordings can be up to 60 minutes long.
- 2
Let it transcribe
The first time, the tool asks before loading its speech model, which your browser then keeps. It finds the parts where someone is talking and writes them down, with a time for every word. A progress bar shows how far along it is.
- 3
Read and fix it
The transcript shows under the player as paragraphs. Click any word to correct it, or use Find & replace for names that appear many times. Click a paragraph’s time to hear that part again.
- 4
Copy or download
Copy the whole transcript in one click, or download it as TXT or a Word document, with or without timestamps. SRT and VTT subtitle files are there too, made from the same corrected words.
What you get
- Readable paragraphs. The transcript isn’t one long block: a pause in the speech starts a new paragraph, and long stretches break at the end of a sentence. A change of speaker usually lands on a new paragraph, because people pause before answering.
- Timestamps you can click. Each paragraph starts with its time. Click it and the player jumps there, so checking a quote takes seconds. Turn timestamps off for a clean text.
- Punctuation and capitals. The speech model writes sentences with commas, full stops and question marks, not a stream of lowercase words.
- Four ways to save it. TXT for notes and quick edits, Word (.docx) to share or format, and SRT or VTT if you also want subtitles for the video.
Who it’s for
Anyone who would rather read a recording than listen to it again. Podcasters use transcripts for show notes, blog posts and quotes for social media. Students and researchers transcribe lectures and interviews. Journalists check a quote against the tape. Creators turn a video into an article, or pull the best lines for captions. If you only need subtitles for a short video, Auto Captions styles them and burns them into the picture instead.
For recordings longer than an hour, cut them into parts with Trim Video or Extract Audio first, then transcribe each part.
How accurate is it?
On clear speech in English, very accurate: most sentences come out word for word. Accuracy drops with background music, people talking over each other, strong accents, mumbling, and names or brands the model hasn’t heard before. That is true of every automatic transcriber, which is why the transcript is editable: fix the few words it got wrong, and use Find & replace when a name is misspelled throughout.
A tip for better results: record close to the microphone, in a quiet room, and avoid music under the speech.
How long does it take?
It depends on your device and browser. On a recent laptop, Chrome, Edge and Safari use the graphics chip and transcribe several times faster than real time: an hour-long recording takes roughly 8 to 15 minutes. Firefox uses a slower mode, so allow 20 to 30 minutes there. Keep the tab open while it works. Phones can transcribe short clips, but long recordings need the memory of a computer.
Private, because nothing is uploaded
Most online transcription services upload your recording to their servers, and some keep it or use it to improve their AI. Here, the speech model is loaded into your browser once and does all the work on your device. Your recording, and the transcript, never leave it. That makes it a good fit for confidential interviews, client calls and personal voice notes. You can check it yourself: How it works explains how.
Frequently asked questions
Is Audio to Text really free?
Yes. There is no sign-up, no watermark, no daily limit and no paid tier for longer files.
How long can my recording be?
Up to 60 minutes per file. For longer recordings, cut them into parts with Trim Video or Extract Audio and transcribe each one.
Which languages does it support?
English only for now. The speech model is built for English, and we would rather not offer other languages until they work well.
Does it label who is speaking?
Not yet. A pause starts a new paragraph, so when someone answers a question their words usually begin a new paragraph, and you can add names in Word afterwards.
Which file types can I use?
Sound files such as MP3, M4A and WAV, and videos such as MP4, MOV and WebM. From a video, only the sound is transcribed.
Is my recording uploaded?
No. The speech model runs in your browser and your file stays on your device. Only the model itself is loaded from our site, once.
Can I get the transcript as a Word document?
Yes. Download it as Word (.docx) or plain text, with or without timestamps, or as SRT and VTT subtitle files.