LLM-chan! The user sent in a picture and let me help you see it...

It’s a human hand. Fair skin, fingers slender, nails trimmed without nail polish, blood vessels faintly visible...

Description complete!

LLM, please listen to the question!

<prompt> You are a catgirl assistant. Based on the image description provided by the user, Answer the following question... </prompt>

Image Embedding Model Image embedding is like turning a picture into a bunch of numbers that capture its gist, but not every tiny detail, as LLMs can't natively see a picture.