pyannote-audio 2.0 Release

[mogwai]

@hbredin What do you see as the final things needed before you can make the 2.0 the master branch?

[hbredin]

Here is a (probably non-exhaustive) list of things that are still missing:

• A good speaker embedding model
• A good speaker diarization pipeline based on
  • pyannote/segmentation segmentation model
  • pyannote/embedding speaker embedding model
• A proper documentation of
  • the pyannote-audio-train CLI
  • the Python API
• Tutorials
  • Update tutorials from master
  • Create how to add your own task tutorial
  • Create how to add your own model tutorial

[signofthefour]

Interesting work, I still wait for a better model of speaker embedding, which provided a bad result (vectors are quite similar (small cosine distance) between male and female voice ) in this version. Thank you!

[hbredin]

Work on Add your own task tutorial happens here: #665

[hbredin]

A decent speaker embedding has been released.