Model with my own training (probably four or more trainings runs overtime), not just a merge.
A pleasant 2.5D style without uncanny eyes. Very responsive to prompts and no inconveniences like mangled fingers.
It's a model to just generate in a chill way. The most important thing is your creativity
Tip:
Use the following Settings:
Always use Adetailer (Automatic1111 extension) with default settings
Sample: Restart
Steps: 13
Upscale: Latent-Nearest Exact, Denoise strength: 0.6, Resize: 1.5
Description
FAQ
Comments (5)
Your model would probably get more attention if you:
Upload a pruned version
Add more example pictures
Add generation data to the pictures
Just some advice, I myself is hesitant to try it out because of the lack of those things.
Okay, I don't care about attention though.
Model is good enough to speak for itself
How many images did you use for training and how many steps on them? Custom tagged?
Different for each training. What has worked well for my last training are these settings. And custom tagged, yes. I use the Hollowstrawberry colab notebooks training_config (1).toml ~ pixeldrain
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.