- Model introduction: portrait, brush strokes, animation style;
- Creative overview: Some characters' material is used, and these characters are labeled. You can try these labels:
Aquaman, black Widow widow, bulbasaur, Catwoman, Charizard, Goku Son_Goku, Sakura Sakura, Kakashi Kakashi, Sasuke Kakashi, Jiraiya, Naruto Naruto, Chunli, Thor Thor, spiderman, Pikachu, Captain America, Iron Man, Wonder Woman, Superman;
You can also use the above two tags or more than two tags mixed, adjust the weight of the tag can be!
- Generation suggestion: the use of animation bottom mold more painting feeling, 3d and realistic also have a different feeling, lora weight 0.6-0.7; If the HD repair is on 2x, the redraw amplitude is recommended to be 0.4, and 0.7 is not on.
Description
- Model introduction: portrait, brush strokes, animation style;
- Creative overview: Some characters' material is used, and these characters are labeled. You can try these labels:
Aquaman, black Widow widow, bulbasaur, Catwoman, Charizard, Goku Son_Goku, Sakura Sakura, Kakashi Kakashi, Sasuke Kakashi, Jiraiya, Naruto Naruto, Chunli, Thor Thor, spiderman, Pikachu, Captain America, Iron Man, Wonder Woman, Superman;
You can also use the above two tags or more than two tags mixed, adjust the weight of the tag can be!
- Generation suggestion: the use of animation bottom mold more painting feeling, 3d and realistic also have a different feeling, lora weight 0.6-0.7; If the HD repair is on 2x, the redraw amplitude is recommended to be 0.4, and 0.7 is not on.
FAQ
Comments (2)
This produces incredible results for characters with long hair and white skin tone - however, at any weight past ~0.5 it's basically impossible to produce something like "african american man with a buzzcut" when using high-res fix (although it seems to perform "ok" on this task without high-res fix). Would be awesome to get a V2 eventually with some variation in hair styles and skin tones / features, as the style it produces (especially on background detail is amazing)
Anything with Niji in it's name is a chinese art reference in which case much of the art will be lighter skin tones. It is the result of the training data. The lora cannot comprehend what was not tagged in it as you hit higher weights you get results closer to what was diffused from the original work. mix it with another Lora or use a checkpoint trained and tagged with things like "african american man" and "Buzzcut" to get the results you need.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.