So I am just goofing around with Stable Diffusion, and opted to get Gemini (formally Bard) to generate some prompts. It is doing fairly well, then I realized Gemini is multimodal. So I ask it to create a prompt based on a picture. Of course, it come back saying that because its " Google's AI Principles to return content that depicts people." Not mad or upset. It is the stance that they are taking. BUT... I do have an RTX 4090 at my disposal and can build my own Gemini.

With that being said, is there any OS multimodal LLMs out there that I can start messing with to get the results I am looking for?

​

(FYI - I am looking to build my own, not looking for some "site" to do the work for me. Only so many subscriptions I can pay for.)