
Style based on artist Azazel1944's works, specifically on the 79 earlier warcraft-era works, though every other image up to the present day (111 in total) has also been tagged and is included separately in the dataset in case anyone wants to play with that.
Model is a LoCon trained on NoobAI ε-pred 1.0. Untested on base ILXL.
Usage
Basic
Works best at a strength of 1.0
, but can be lowered to 0.8
if you can't get rid of the borders.
Borders are tagged to be included/excluded, with different subtypes:
-
main tag:
border
-
frame border:
picture frame
-
letterboxed border:
letterbox
-
Sub themes
While this is primarily warcraft
work, there are also other of the artist's focuses, including asian
girls and warhammer
.
Issues
It seems that even when trying to overtrain this, the style doesn't come out quite as accurately as I'd hoped, but you can try adding traditional media
in some cases to nudge it a little closer.
Extra
In the interest of openness and community-helping, I have uploaded the dataset along with the model, this includes: images, captions, and the training json (for kohya_ss).
As noted in the intro, there is also a separate folder in the dataset including fully-tagged images which were not part of the model's training.
描述:
-
trained on 79 images on NoobAI ε-pred 1.0
trained on 79 images on NoobAI ε-pred 1.0
训练词语:
名称: Azazel1944_NAIXL_v10.dataset.zip
大小 (KB): 43152
类型: Training Data
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
名称: Azazel1944_NAIXL_v10.safetensors
大小 (KB): 114314
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success