Type | |
Stats | 532 1,169 |
Reviews | (73) |
Published | Jun 6, 2023 |
Base Model | |
Training | Steps: 1,750 Epochs: 10 |
Trigger Words | PowerScreencap |
Hash | AutoV2 56F6E4F15A |
Please read the description below!
This is an experimental Character LoRA that was trained with only anime screencaps in training data on a new specific model that could be used to emulate specific anime styles, with the use of supplementary Anime Style models.
This is not a generic character model, even though it could in theory be used as one.
I will release a generic Power Character LoRA trained on AnyLora soon to compare the results, so stick around for that :) (Both models contain the same screencaps, but the other one also uses fanart images in training dataset)
Big thanks to Lykon (https://civitai.com/user/Lykon) for making this possible!
To achieve similar results as showcased, it is necessary to follow a few simple steps:
Use AnyLoRA Anime Mix Screencap Style Model to generate images (https://civitai.com/models/84586/aam-anylora-anime-mix-anime-screencap-style-model).
Use Chainsaw Man Anime Style Lora (https://civitai.com/models/53149/chainsaw-man-anime-style-lora) with the weight of 0.4-0.6.
Pay attention to removing certain tags that may disrupt the style (CGI, photorealistic, RAW photo, e.t.c.)
Recommended weight for this LoRA: 0.7-0.85
Activation word for this LoRA: PowerScreencap
Use of " beautiful detailed eyes, ", or other eye enhancement LoRA's is not recommended for this Character LoRA.
Use of "Detail Tweaker LoRA" model is recommended for better details.
Upscale to 1.5.
I haven't received the same quality of results using other anime SD checkpoints, but if you want to use it as a generic character LoRA make sure to lower the weights.
I would appreciate any reviews, positive or negative!
I would very much appreciate you sharing your results ^_^
If you have any suggestions on what LoRA i should do next, please share them in the comments
Enjoy :)