There's a highly nontrivial amount of creative work that goes into getting high quality stuff out of the diffusion models. One-step txt2img is a parlor trick.

There's a highly nontrivial amount of creative work that goes into getting high quality stuff out of the diffusion models. One-step txt2img is a parlor trick. Those of us who are seriously using these tools are often making many round trips to/from GIMP and img2img, txt2img, doing compositing, painting and masking. It's definitely a process that stretches creative muscle. Check out some of InvokeAI's canvas overdrive videos [0] for a bit more of a sense of how an artist works with that implementation. [0] https://youtube.com/@invokeai https://youtube.com/@invokeai