V1.0-A
Hello everyone! This last merge really surprised me with its versatility, I hope you use it and find new ways to achieve interesting and beautiful things!
So far I have found 2 interesting ways to use it, the first is the typical one with hard prompting and using the negative prompt, this will give quite clean results and very focused on the prompt, the second is not using the negative prompt, giving rise to a more chaotic scenario where some generations will not be the desired ones but some will be gems!
I insist if you have different and creative ways to try it, upload your contributions!
Recommendations
STEPS: 20-30
CFG: 5-6
DPM 2M Karras or Euler a
!!!!it's experimental at the moment!!!!!
V0.7-A
works best without negative prompt, or just
lowres, red eyes,feel free to experiment and publish your results, so far this model has given me a lot of entertainment!
Recommendations
STEPS: 20-30
CFG: 5-6
DPM 2M Karras or Euler a
V0.6-A
At the moment this model is my favorite for getting concepts, it is incredibly responsive with a very vague prompt it can give you very good abstractions of very vague words!
use it without negative prompt!!
Description
Crazy improvement in this version, try it!
FAQ
Comments (14)
This model is my favorite model. It doesn't work well on noobai fooocus but this model works perfectly. I wish you continued success. I'm happy that the new version is out.
I really appreciate your comment! I will continue trying to find new creative ways to take advantage of these models, and I will also be uploading when I achieve something!
"Crazy improvement in this version"
So what have you changed? We don't have your Info and saying that doesn't help much.
There are nowadays so many Models with 0 Info wich get ignored because lazy description.
Does this model require a VAE? When using the SDXL VAE, it reduces the saturation of the image. However, if I don’t use a VAE, the inpainted areas during partial redrawing show noticeably higher saturation compared to other regions.
I generally use the same vae of the model in the nodes of the checkpoint load in ConfyUi but in other occasions I use a separate vae, the "xlVAEC_f2" (https://civitai.com/models/152040?modelVersionId=719135) you can try it and tell me if that solved your problem!
@A21_ Thank you. Compared to not using VEA, this results in a smaller saturation deviation, but the problem still persists.
@cangtianbaiqi What program do you use for your img? (Confy, a1111, forge, foocus, etc.)
@A21_ I use forge
@cangtianbaiqi Sorry for the delay! I installed Forge and tried to explain but it works correctly (using the vae) I don't know how to fix it, sorry!
@A21_ Thanks. It seems that each img-to-img process makes the colors more vibrant. I will try to minimize the number of img-to-img iterations to avoid this issue.
I really love this model, but has it stopped receiving updates?
I'm still quietly hoping for further improvements in body stability and background rendering.
hello thank you very much for the support! i really moved to something a little different with a model i've been testing for months, i'm planning to write an article if you are interested in it!