Transcribe audio

Turning a recording into text

Upload an audio file or a video and in a moment you have the text. It suits meeting minutes, a lecture, an interview and the voice note you would rather not type out by hand. Besides the running text you can download subtitles as SRT or VTT and attach them straight to a video.

The model understands English, and German and Polish as well. When you know the language of the recording, set it: the result is usually more accurate than when the model has to guess.

What affects the quality of a transcript

• How close the microphone is. A phone lying on a table in the middle of the room does worse than a recorder in front of the speaker.
• Background noise. Music, traffic and a humming air conditioner cost a human listener clarity, let alone a model.
• How many people talk at once. Interruptions are the hardest thing for a transcript. The tool also does not tell speakers apart, so the text carries no names.
• Technical terms and names. Abbreviations, foreign surnames and numbers are where a transcript most often goes wrong. Those are the parts to check.

Where the recording is processed

For this tool, unusually, not in your browser. The speech recognition model is too big for that, so our server does the work. The recording therefore leaves your machine, is processed in memory and disappears once it is done. We do not store it, archive it or pass it to anyone.

If it is a sensitive recording that must not leave your computer, this is not the tool for it.

Frequently asked questions

How long a recording can I upload?

Up to ten minutes and 120 MB. Split a longer recording first, with the audio trimmer for instance. The limit keeps the processing time reasonable: the server has two cores and computes for everyone at once.

How long does a transcript take?

Roughly a sixth of the length of the recording. A ten-minute recording is usually done within two minutes. Leave the page open while it runs; the result appears by itself.

How many times a day can I use it?

Once a day free, twenty times with a Premium account. Transcription is the most computationally demanding thing on the whole site, hence the stricter limit than the other tools have.

Which files does it take?

MP3, WAV, M4A, AAC, OGG, OPUS, FLAC, and MP4, MOV or WebM from video. With video, only the audio track is used.

Does it tell who is speaking?

No, it does not do speaker separation. You get the text in timed sections, but with no note of who said what.

What are SRT subtitles for?

You attach them to a video on YouTube, in a player or in an editing program. The file holds sections of text with the times they should appear.