Diarization (beta)

• macos (Finder)
  1. Open a new Finder window and choose Go > Go to Folder... from the window menu
  2. Type ~/Library/Application Support/Phon in the dialog and press Enter
  3. Create a new folder plugins if this folder does not already exist
  4. Copy the downloaded file, phon-diarization-plugin-<version>.jar into the plugins folder.
  5. Restart Phon
• macos (command line)
  Copy and paste the following into a new Terminal window (create destination folder if it does not exist:)

  cp ~/Downloads/phon-diarization-plugin-<version>.jar ~/Library/Application\ Support/Phon/plugins/

• windows
  1. Open a new Explorer window and type %APPDATA%/Phon into the address bar
  2. Create a new folder plugins if this folder does not already exist
  3. Copy the downloaded file, phon-diarization-plugin-<version>.jar into the plugins folder.
  4. Restart Phon

The LIUM diarization tool (https://projets-lium.univ-lemans.fr/spkdiarization/) is an embedded (i.e., offline) diarization tool which is packaged with Phon.

• Max speakers

  Results must conain no more than the specified number of speakers. A value of 0 indicates this parameter is not considered.

Requires a Google Cloud account and Internet access. Audio for the session will be uploaded and transcribed using Google Speech-to-Text. A Google Cloud account is required to use this function. See Setting Up Your Google Cloud Account for instructions on setting up Google Cloud for use with the diarization tool.

• Project id

  The project id for your Google Cloud project available from the IAM > Settings page in your Google Cloud account page.

• Google Cloud service account credentials file

  The service account credentials file (.json) created when Setting Up Your Google Cloud Account.

• Google Cloud Storage bucket location

  The storage bucket location if the audio for diarization is longer than 60s. For best results choose the option closest to your geographical location.

• Language model

  The language model used for Speech-to-Text transcription. (Only languages which have the diarization feature available are shown the list, more information can be found at https://cloud.google.com/speech-to-text/docs/languages.)

• Audio format model

  The format of the audio file. Choose the format most appropriate (for most people this will be video)

  • default

    Use this model if your audio does not fit any of the other models. For example, you can use this for long-form audio recordings that feature a single speaker only. The default model will produce transcription results for any type of audio, including audio such as video clips that has a separate model specifically tailored to it. However, recognizing video clip audio using the default model would like yield lower-quality results than using the video model. Ideally, the audio is high-fidelity, recorded at 16,000Hz or greater sampling rate.

  • video

    Use this model for transcribing audio from video clips or other sources (such as podcasts) that have multiple speakers. This model is also often the best choice for audio that was recorded with a high-quality microphone or that has lots of background noise. For best results, provide audio recorded at 16,000Hz or greater sampling rate.

  • phone_call

    Use this model for transcribing audio from a phone call. Typically, phone audio is recorded at 8,000Hz sampling rate.

  • command_and_search

    Use this model for transcribing shorter audio clips. Some examples include voice commands or voice search.

• Max speakers

  Results must conain no more than the specified number of speakers. A value of 0 indicates this parameter is not considered.