Any support for image generation models? (e.g. Stable Diffusion) #3056
-
Hi there! I've been using Jan for quite a while and it's become my main chat app at this point. I was wondering though if there was any support at all for image generation models such as Stable Diffusion. I didn't see anything in the docs and I can't seem to find generation models in the hub. So, are these supported? Or is the support coming sooner or later? Or is it an explicit non-goal of the project? Thanks! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
at the moment Jan dosnt support ( i dont now if its a goal) try koboldcpp for that ;) for documents still GPT4ALL is best also for other quantisation |
Beta Was this translation helpful? Give feedback.
at the moment Jan dosnt support ( i dont now if its a goal)
try koboldcpp for that ;)
for me the image generation is too complex (like a1111) to merge it in an LLM software, if than is a link the best so that the image is called by a prompt from LLM . . .
for documents still GPT4ALL is best also for other quantisation