Skip to main content

Speaker assignment in transcript

Feature Request

Ability to have this would be extremely powerful - even if it’s just Speaker 1, Speaker 2, Speaker 3 etc and then provide the ability to updat ethe speaker “names” for the recording and be able to correct it and reassign it if needed eg set one piece of the recording / transcript Speaker 1 —> Speaker 2

There was a diarisation capability on the roadmap that suggests this was completed but no evidence of this in the application against the transcripts so may be there but requires better visibility?

Posted by ajfisher ·
Status: Completed5 comments

Log in to comment and vote

Comments5

  • Vowen

    Team•

    Apr 14

    @ajfisher We have speaker diarization on models that support it at the moment:

    You should see this option before you start to take notes.

    And after the notes are created, in the transcription section of the notes, in the right panel you should see this where you can map the speaker names.

    Right now the models that support diarization are: Mistral, AssemblyAI, Deepgram, and ElevenLabs. We’re working on adding support for it through Parakeet local model as well.

    • ajfisher

      •

      Apr 14

      Thanks @Vowen - any specific Mistral models needed to make this work? I’m assuming this is because I’m using the OpenAI whisper local models??

      • Vowen

        Team•

        Apr 14

        Yes, Local open AI models currently don’t support diarization. Though we are also looking for ways in getting diarization in such models as well. For Mistral, you should be able to get API keys from their console dashboard. They offer free API Keys. Thanks

    • Georgii Vlasov

      •

      May 9

      This function does not work when exporting; speaker names are lost in file.

      • Vowen

        Team•

        May 9

        Thanks for flagging this @Georgii Vlasov , we’ll get it fixed in the coming update.