To transcribe audio to text, you need a speech recognition tool, a reasonably clear recording and a few minutes to check the result. The free Audio to Text tool on this site does the recognition in your browser, so the file stays on your device, and it saves the result as plain text or as subtitles with timings. Plan to read the transcript once: automatic transcription is good, but never perfect.
What you need
- An audio file (MP3, WAV, M4A and other common formats) or a video file. The sound track of a video is used directly, so there is no need to extract it first.
- A computer or a recent phone. A computer is faster for long recordings.
- A short one-time download the first time you use the tool, while the speech engine is loaded. Your browser keeps it for next time.
Step by step
1. Open the recording
Go to Audio to Text and choose your file, or drop it on the page. After a moment you see its length and size, and a prompt to choose the language.
2. Choose the language
The Language list decides which engine is used:
- English (fast) is the smallest and quickest. It suits clear English speech, such as a podcast or a dictated note.
- English (more accurate) is slower but copes better with accents, background noise and several speakers.
- Detect the language works it out for you, which helps with unfamiliar or mixed recordings.
- Or choose the spoken language directly. Around 100 languages are listed, including Spanish, French, German, Chinese, Japanese, Arabic and Hindi.
If you pick anything other than the two English options, a Translate to English box appears. Tick it if you want an English translation instead of a transcript in the original language.
3. Click Transcribe
Text appears as it is recognised, so you can start reading before it finishes. On a recent computer, an hour of English audio takes roughly one to three minutes with the fast engine, and a few times longer with the multi-language one. Keep the tab open. Cancel stops the job if you picked the wrong option.
4. Correct the text
The transcript is in an editable box. Fix names, technical terms and anything the engine misheard. If you plan to save subtitles, your corrections are kept with their timings.
5. Save or copy it
Under Download as, choose:
- Text (.txt) for notes, articles, meeting minutes or quotes.
- Subtitles (.srt) for YouTube, VLC and most video editors.
- Web subtitles (.vtt) for HTML5 video players on websites.
Then click Download, or Copy text to paste it into a document or email. If you mainly need captions, the Subtitle Generator page opens the same tool with SRT already selected.
Tips for better results
- Start with good audio. A phone placed close to the speaker beats an expensive microphone across the room. Background music, echo and people talking over each other cause most errors.
- Pick the language yourself when you know it. Detection is reliable on clear speech, but choosing the language avoids a wrong guess on short clips.
- Use the more accurate English option for difficult recordings. Phone calls, strong accents and noisy rooms are where the extra time pays off.
- Split very long recordings if your device is slow or low on memory, for example a full-day workshop. Transcribing in parts is easier to check, too.
- Check numbers, names and quotes against the audio before you publish them. Spelling of people’s names and brand names is where any automatic transcript is weakest.
Common problems and fixes
“No speech was found in this file”
The file may be silent, extremely quiet, or contain only music. Play it to make sure there is speech, and check that you chose the right file rather than an empty export.
The text is in the wrong language or looks garbled
Detection guessed wrong, which happens on short clips or when the first seconds are music. Choose the spoken language from the list and run it again.
Some words are consistently wrong
Unusual names, jargon and abbreviations are often misheard. Fix the first one, then use your text editor’s find-and-replace for the rest after downloading.
The file will not open
Very unusual formats or damaged files may not be readable. Convert the file to MP3 or WAV with an audio converter and try again.
Alternatives
There are several other good ways to get a transcript, depending on what you already use:
- Built-in phone features. Recent iPhones show a transcript for recordings in Voice Memos, and many Android phones have a recorder app with transcription. These are handy for recordings you make on the phone itself.
- Office suites. Microsoft Word for the web has a Transcribe feature for uploaded recordings with a Microsoft 365 subscription, and Google Docs offers live voice typing (it listens to the microphone rather than reading a file).
- Video platforms. YouTube creates automatic captions for uploaded videos, which you can edit in YouTube Studio.
- Paid services. Human transcription services are slower and cost money, but they are the right choice when accuracy must be near perfect, such as for legal or medical records.
FAQ
Is transcribing audio this way private?
Yes. The speech is recognised on your own device, in the browser, and the recording is not uploaded. That makes it suitable for interviews, meetings and voice notes you would rather not send to a server.
How accurate is automatic transcription?
On clear speech from one speaker, most words come out right. Accuracy drops with noise, overlapping voices, strong accents and specialised vocabulary. Always read the transcript before relying on it.
Can it tell who is speaking?
No. The transcript is one continuous text with timings, without speaker labels. For an interview, add the names yourself while you read through it.
Can I transcribe a video?
Yes. Choose the video file directly. Only its sound is used, and you can save the result as subtitles to add back to the video.
Is there a length limit?
There is no fixed limit. Long recordings simply take longer, and very long ones may need more memory than an older phone has.