1.0a is harder to control than 0.6t, but can more and do sharper results average without img2img or hiresfix.
use 0.6t if you want stable results on easy scenes.
Bakuen/konosuba models will be updated later
and i think i made result LoRa's page after it's finished/or get close to it
Works fine with anime models like andromedaMix or grapefruit or https://civarchive.com/models/4669/corneos-7th-heaven-mix
strength 0.6-0.8, hiresfix or img2img for best results(it's works for every LoRa).
Train models isn't easy, crypto donations on some coffee are welcome:
Description
little updated Megumin/bakuen model, i will do many fixes in next 10days about it, but it's possible to use and get different styles from anime with it at now too.
0.6t model easier to control at now. 1.0a harder to use and need more precisious prompt->if you want just easy scene or something like it use 0.6t.
strength 0.6-0.8, hiresfix or img2img for best results(it's works for every LoRa).
if you get bad results - try https://civitai.com/models/4669/corneos-7th-heaven-mix this checkpoint and write in prompt position(like "front view, face"), because dataset not only front-face.
Train models isn't easy, crypto donations on some coffee are welcome:
FAQ
Comments (2)
The style is faithful to the anime. Unlike my loras, which come out a bit different, with more detail. It would be interesting if you shared a bit of your wisdom
answer is pretty easy, if you train on arts you get arts, if you train on anime you get anime. (ofc with some prompts(just copypaste from my images and mutate from it as start point) and need to balance it correctly, but it's details you get from experience).
P.S. ofc choise base model on train LoRA with knowns booru tags(anythingv5?, my model, 100500 other anime models), because if you choise SD 1.5 you get worse results.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.