Clip/Blip - Something I know very little about

[myndxero]

I get the basic scenario that these bridge the gap between user prompt and generated result. What I'm mostly curious about, while viewing Vlad's wiki here, makes mention of using other CLIP models and such. I was wondering what people's experiences are with this and if there were any models they might recommend?

Also, since a lot of information on this seems fairly dated, it's unclear to me whether or not you have to train the model on the specific CLIP/BLIP model still or if that's not the case anymore if they're interchangeable.

Thanks!

EDIT: poked my head in here, so many and I don't know enough to look through them with any degree of understanding:
https://huggingface.co/models?other=clip