An experimental command line application for transcribing audio files, written in Swift. Supports macOS Catalina (and above?) only.
$ ./transcriber --help
Usage:
$ ./transcriber
Options:
--list-locales [default: false] - Print the set of supported locales (then exit)
--file [default: ] - Audio file to be transcribed
--locale [default: en_NZ] - Locale of the speakers (use --list-locales for options)
$ ./transcriber --file sample.mp3
Recognising audio. This will take about as long as the length of the clip...
This is a sample sentence
Assuming you have XCode installed on your Mac, swift build should do what you need.
- Requires macOS Catalina
- Apple's restrictions on speech recognition are not well-documented, but you might find you are rejected if you make too many requests.
- Clips longer than 1 minute will return a truncated transcript.
- I couldn't get offline recognition to work (all locales report that it is not supported).
If you can't get it to work, try these things:
- Check your internet connection.
- Ensure you have authorised Speech Recognition if macOS asks you.
- Ensure Siri is enabled in macOS Settings.
- Try a shorter audio file (usually anything longer than 1 minute is rejected or truncated).
- Wait and try again - Apple limits the number of server requests per hour.
- Use a macOS supported audio file format (aac, wav, mp3 etc).