Roleplay in 2026
Posted by No_Afternoon_4260@reddit | LocalLLaMA | View on Reddit | 8 comments
hey, not my kind of topic usually.
looking for a framework or something to generate illustrated stories for kids.
it's got to be stateless (serverless) the llm endpoint is local but the image gen got to be api (no resources to allocate for it). is there anyway to get character consistency across images without some over engineered comfy workflow?
UnbeliebteMeinung@reddit
Sillytavern
Status_Record_1839@reddit
fal.ai has a character consistency API that's stateless and cheap enough for kids' story volumes, might be worth a look before building your own pipeline
No_Afternoon_4260@reddit (OP)
Great thx a lot
No_Afternoon_4260@reddit (OP)
Great thx a lot
jacek2023@reddit
I think ComfyUI requirements are lower than LLM requirements
No_Afternoon_4260@reddit (OP)
Yes I understand but we don't want the cold start and we don't have ressources for that thing to idle all day long.
Geritas@reddit
You can use an api within comfy workflow instead of local generation. Some image editing model with a character ref would work fine. Though I heard it can get somewhat expensive, idk, never used apis in comfy.
No_Afternoon_4260@reddit (OP)
That's true forgot about that, thx