LesskeysAudio to text, without the upload, on Mac or Windows
To turn audio into text, add the file to Lesskeys: it transcribes it on your computer, labels the speakers and uploads nothing. It keeps working with the internet off.
macOS 13+ on Apple Silicon · No subscriptionWindows 10 and 11 · No subscription
Last reviewed
Drop it in
How to turn audio into text
- Open Lesskeys, choose Transcribe and add the file.
- Leave the language on auto-detect, or pin up to four.
- You set how many people spoke (2 by default).Say how many people spoke: pick 1 to 4, or enter a custom number.
- Start it. Read the text in Library, press play on a line to play from there, or Copy all.
The first run downloads a small set of speaker files.
Who said what
A speaker label on every line
Every line comes with a speaker label.
Rename Speaker 1 to a real name, or merge two labels if one person was split. A named voice can be recognised in later recordings; the voiceprints stay on your computer. Speaker labels covers what makes labels wrong.
Your computer
The file never leaves your computer
A converter website uploads your file. Lesskeys reads it from your disk and writes the text next to it.
- Nothinguploaded
- Offlineworks with the internet off
- 99languages
Switch Wi-Fi off and try: transcribe a video without uploading it.
Check it
When an audio transcript is wrong
It is only as good as the recording.
- Noise and distance
- Names and jargon
- Two people at once
- A long file can stop early
Check names and technical terms. On a long file the speech model can drop the end and still say it finished: read the last lines against the audio (when Whisper stops early). Where a person must check every word, use a service staffed by people. What is on-device transcription has our accuracy figures.
On a Mac
Which files work on a Mac, and how fast
Audio or video, no conversion for the common types.
- Audio: MP3, M4A, WAV, AIFF
- Video: MP4, MOV, M4V
Other formats may not open: convert them first. In our test on one Apple M1 Max, an hour of audio took about five minutes with the more accurate model. Finding the speakers adds time, and an older Mac is slower.
Source: Which Whisper model to run on a Mac.
On Windows
Which files work on Windows
Audio or video: the picker accepts the common types.
- Audio: WAV, MP3, M4A, FLAC, OGG, WMA, AIFF
- Video: MP4, MOV, M4V, MKV, AVI, WEBM, WMV
Whether a particular file opens depends on how it was encoded: if one will not, convert it to MP4 or M4A first. Windows speed has not been measured here, so there is no figure for it. The first transcription downloads a small set of speaker models. No file yet? Record a voice note on Windows first.
Questions
Questions about converting audio to text
How do I convert audio to text without uploading it?
Use an on-device transcription app. Lesskeys transcribes audio and video files on your computer: the file is read from your disk, the text is written next to it, and nothing is uploaded. No account is needed.
What file formats work?
On a Mac: audio MP3, M4A, WAV and AIFF; video MP4, MOV and M4V. Other formats, such as MKV, WEBM, WMA or AVI, may not open there; convert them to MP4 or M4A first. On Windows the picker accepts WAV, MP3, M4A, FLAC, OGG, WMA, AIFF and common video types; convert a file that will not open.
Can I transcribe a long recording, such as an hour-long meeting?
Yes, an hour-long meeting is a normal job. A long recording takes longer to transcribe than a short clip, and a slower computer takes longer still. Read the last lines of the transcript against the audio to check that nothing was cut off.
How accurate is it?
It depends on the recording more than on the computer. In our test on 6,086 words of public audio, the most accurate speech model got about 2 words in every 100 wrong. We did not test cloud converters, so we do not claim a match: try your own audio.
More