Honesty first as some may know this was the first lora I made and as you can tell by the skipping version numbers, it's always hated me. Though before beginning this one is a LoCon unlike my previous lora (seems to have helped a bit). Why it doesn't show up as such I don't know.
This is a model that's based on the monster girl feature of snake hair. Tends to work best at 0.6. This model was trained on NAI anime-full-pruned and generated in Anylora. Made with the Kohya tool for AI Training through google colab. Any suggestions to improve or recommended things for the data set would be appreciated.
Description
First version that actual makes recognizable snake hair.
This was a test to learn how to use the Kohya tool for AI Training. The model was trained with a mix of snake hair pictures as well as actual snakes. This model should not make characters with a mix of snake and real hair in favor of just snake hair. Any suggestions would be appreciated. Lora works best in the 0.6-0.7 range. This model was trained on andite's anything 4.5 from huggingface.
FAQ
Comments (4)
Very interesting idea, thank you for sharing.
great idea. Might be good to try a triggerword that isn't snake. snakehair together might work, or Medusa... and non anime sources to keep lora neutral
Not really sure about the non-anime sources but the new update does have a single phrase trigger word.
@rulles did it help with extra snakes? personally i always struggle with descriptors that are "strong". 'eagle beak' can give wings and 'boar snout' can change a human into a pig...
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.