8 Building Advanced Workflows with ComfyUI
This chapter covers
- Introducing ComfyUI.
- How to implement a basic diffusion workflow.
- Extending basic workflow to support “hi-res” upscaling.
- Using existing ComfyUI workflows.
- Repeatedly running an img2img upscaling workflow.
“When you make something easier, you may lose your edge in a skill. Convenience and ease aren't the same thing.”
--Shenshi, Dungeon Meshi (aka Delicious in Dungeon)
Careful readers might have realized that we ended the last chapter in a position which contradicts one of the more important pieces of advice in the entire book: generate lots of images!
The workflow we used for A1111 to create an upscaled image using the img2img capabilities of Stable Diffusion was powerful but painful to reproduce for multiple images. For each image generation you are required to manually repeat multiple steps (we’ll refresh on this in a bit). This makes it very tedious to produce 10 or 20 sample images of your upscaling.
In this chapter we’ll be learning about a much more powerful tool, ComfyUI, that will enable us to scale our image generation to more complex tasks while still being able to easily generate as many images as we’d like. When new to Stable Diffusion, ComfyUI can be quite tricky to master (which is why it’s best to start by learning A1111), but now that we understand the details of how Stable Diffusion does image generation, it will be much easier to get started!