I do a little tinkering before posting, but the goal is to see how well the models can actually understand what's going on in my smutty lewd images, so my workflow is mostly messing with the system prompt rn.

I've spent a bunch of time playing with local Stable Diffusion images w/ LoRAs trained on me... and this new experiment involves feeding those pix into local & uncensored vision-capable LLMs with instructions to "write a story that goes with this picture".

Yaaaay thanks & hi!! ✨