I'm trying to understand how someone can run a specific ControlNet and Stable Diffusion model with scalable GPU resources. How would someone design a system to achieve this? I've messed around with models on Replicate, but none seem to do a good job with converting a doodle to a photorealistic image. I can do it perfectly fine in the Stable Diffusion web UI, but the API for that is only accessible locally. Anyone have any ideas or can guide me in the right direction for building a "server" to do this?
[link] [comments]