New models of local transcription

DEVONthink Artificial Intelligence
1

Please allow us to add transcription to the searchable text or add other local transcription models (Whisper, Parakeet).

Problem:

  • I want to transcribe locally using Whisper or Parakeet, but only Apple models are available. The quality of Apple’s LLM does not suit me.

  • I wrote an Apple Script and transcribe audio locally using Parakeet.

  • I can save the transcription only in an annotation, comment, or any text file linked to the media file by a link. I can’t add it to the Searchable Text.

  • If the transcription is not added to the Searchable Text, then automatic tag assignment, sammari and other AI features do not work, since the annotation is not considered equal to the parent file.

1 Like
2

Thank you for the suggestion, an upcoming release might support setting the plain text AppleScript property of images & audio/video files.

3 Likes
3

Agree. Apple’s built-in transcription API is very convenient but not great quality. Would love to use Voxtral locally for private transcription instead. Just as I can use Ollama for chat locally (thank you!), I’d like to continue to expand the services I can use locally.