Please help me do more LORA's by tipping Buzz on the model or by posting images of it I'd greatly appreciate it c:
REQUIRES PIXEL ART EXTENSION FOR A1111 IN ORDER TO HAVE A WORKING PIXEL ART!!
https://github.com/mrreplicart/sd-webui-pixelart.git
It works.... Sometimes...
Not gonna lie it's highly unreliable, out of 20 gens 1-3 pics are passable, still manual fixing is required, that's why it's still on beta. 18 Epochs and a very limited dataset.
It does work along with character LORA's but not always... What I've seen is that female character LORA's tend to work better than male ones with this model.
Might help someone who's using essentials or Idk...
Trigger: gen3backtrainer
Weight: 1
The prompting part is kinda strict tho'. A prompt would go like this...
backsprite, gen3backtrainer, sprite sheet, pixel art, from behind, DESCRIBE YOUR CHARACTER HERE, white background, simple background <lora:PkmnGen3TrainerBack:1>Sampler: I've tested Euler A and DPM++ 2M Karras and there's not a noticeable difference so I'd suggest using Euler A since it's faster
Steps: 24 is the spot for me
Size: 1024 x 1024 (THAT IS IMPORTANT!!! USE 1024 X 1024 IN ORDER TO USE PIXEL ART EXTENSION AND DOWNSCALE IT LATER)
CFG: I've been using it with 5
It's not necesary to add a negative prompt. In fact, it's recommended to not do so.
REQUIRES PIXEL ART EXTENSION FOR A1111 IN ORDER TO HAVE A WORKING PIXEL ART!!
https://github.com/mrreplicart/sd-webui-pixelart.git
You can also check the prompts on the example images.