This does have some usefulness, but personally I am going reverse of this for the moment. I've actually been wondering in my crazy mind if a setup like this might work with an SDXL refiner at the end. Like, just stick a full 1.x with HiResFix right between the base and refiner of an SDXL workflow. Nah, that's crazy... right?

Although, right now, the 1.0 refiner seems broken to me, so I've just been using the XL 1.0 base as a refiner to my 1.x gens. Unfortunately I lost my 0.9 refiner and haven't bothered getting it back. Great work, I will keep an eye on your stuff!

Fantastic! I've had a hard time getting good results in SDXL for photorealistic items. Even after refining, blurry and low res and jacked up teeth, stuff like that. Using this workflow with a 1.5 model as a refiner gives me so much better results, and honestly it's better than what I've gotten with either model on their own. Outstanding work!

Really great to share for people that doesnt know this little trick! It can be done as much time as you want with different models which gives more power to creativity ! :D

Newbie here, can you help explain what type of prompt would go in the 1.5 section and what would go in the sdxl section? I assume you don't use the same pos/neg prompts for both?

Fantastic, I'm a newbie though, and what interests me the most are loras to pin into your workflow. How do you add a stack of loras to your workflow? Namely, I'm interested in SD1.5 compatibile loras to influence the results.

A little confused about how to use this? Can anyone tell me?