search

mobiusml / aana_sdk

Aana SDK is a powerful framework for building AI enabled multimodal applications.
https://www.mobiuslabs.com/
Apache License 2.0
26 stars 3 forks source link

Updated Chat with Video Project #128

Closed movchan74 closed 2 months ago

movchan74 commented 2 months ago

Summary: This PR updates the chat with video project, introduces a new LLM with larger context size, and makes adjustments to dependencies and error handling.

Key Changes:

  1. Removed timeline trimming from the chat with video project.
  2. Added InternLM2.5 7B chat model for use in the chat with video project.
  3. Implemented VideoTooLongException for videos exceeding 20 minutes.
  4. Disabled batched Whisper functionality from our fork temporarily.
  5. Replaced opencv-python with opencv-python-headless to remove libgl dependency.
movchan74 commented 2 months ago

👍🏽 question: What do we want to do about transcribe_in_chunks? keeping them as commented for later?

Keep it for later. Update the code once batched faster whisper is merged into a main repo.