Welcome!
This one is still a work in progress! I'm working on a model that makes a hyper-detailed drawing style, but the vibe of this model just resonates with me. It makes great clothed or unclothed images, good prompt adherence, and some flexibility for moderately complex prompts. I'm pretty happy with the images I've created with it, so I wanted to share it with you all. If you enjoy it, feel free to leave a note or post a gallery!
My Settings
Sampler -DEIS (or Euler A)
Scheduler -SGM Uniform
Steps - 35-45
Resolution -896x1152, 832x1216 (or flip for landscape)
Hires.fix:
Upscaler - 4x_foolhardy_Remacri
Steps - 12-15
Denoise -.3-.4
Upscale by - 1.5 - 2.0
Description
Complete rebuild of the model. Overall it has a better hyperdetailed drawing style with better pigmentation and lighting.
FAQ
Comments (8)
Excellent checkpoint - very strong styles and art - great work!
Thank you so much for the kind words!
I loved Sketch v.1 which was amazing, and i am happy to see that v.3 is not any worse. Excellent quality of picture.
Thank you! I hope you enjoy the new version just as much!
Well, hello there, beautiful.
Hope you enjoy it!
@GhostBear1111 I little less illustrij/niji gothic would be great (if those are what you merged) for facial diversity. You'll note my images all have different hair and skin, but the girls still have similar faces. However, the prompt adherence and chaos of the model makes for very fun images, even with my prompts that are more geared for Chroma/Krea/Qwen.
@clueless_engineer Oddly enough this one has some Jedpoint and a couple of basic anime models that I used to smush LoRAs into. I'll see if I can get the next version to have more face diversity. I appreciate the feedback and the posts!