[bounty] Video LLM for Search

[BenraouaneSoufiane]

Enable more powerful search using visual and audio context.

eg

• [ ] Use Video LLMs like:
  • [ ] Video-LLaMA
  • [ ] Video-ChatGPT
  • [ ] MiniGPT-4 + CLIP
• [ ] Convert video to frames + audio:
  • [ ] ffmpeg to extract frames/audio
• [ ] Send multimodal input to the LLM
• [ ] Output: searchable embeddings or semantic summaries

all the exact things that will need to be done to receive the bounty.

precision is important otherwise the bounty cannot be awarded.

/bounty 400

This is neccesary as matches with user needs

This issue is a response/relied to this issue: #1142

[BenraouaneSoufiane]

@louis030195 I just created the second issue concern this #1142 if you can take a look