ScribeKitCommand line
The scribe command that ships with the package.
scribe is a small front end for trying ScribeKit on real files. Run it from a clone of the repository:
swift run -c release scribe <command>The transcript goes to standard output; progress and a summary go to standard error, so you can redirect the result to a file.
Download the models
swift run -c release scribe prepare # speech model only
swift run -c release scribe prepare --diarize # speech and speaker modelsTranscribe a file
swift run -c release scribe meeting.m4a --lang en --format md > meeting.md| Option | Effect |
|---|---|
--lang <code> | Language hint, for example pl. Detected when left out. |
--no-diarize | Skip speaker detection. |
--speakers <n> | Exact number of speakers. |
--format <f> | md (default), txt, srt, vtt or json. |
Transcribe a call
swift run -c release scribe conversation microphone.wav system.wav --me "Anna" --format srtThe first file is the microphone, the second the system audio. See Call recordings. --me sets the name of the local speaker (default Me); the other options work as above.
At the end, scribe prints the number of words and speakers, the length of the audio and how long it took.