Transcriptions and Chapters

[jj-style]

Podcasting 2.0 supports transcriptions of podcasts and chapters.

I wonder if we could, when a new episode is available (i.e. set up subscribers for the RSS feed APIs):

• transcribe the audio using whisper
• use a text classification model to
  • summarize the whole transcript? For a description of each episode
  • assume chunks of content in 1 hour intervals, summarize each chunk and create chapters for these?

Obviously this would be a fairly slow and expensive process. Probably want to:

• raise event for new episodes
• have transcriber services listening for events
• persist episode information in a database
• still return RSS content and responses if the additional information hasn't been computed yet (i.e. optionally get data from db)