Type | |
Stats | 775 8,627 3.2k |
Reviews | (139) |
Published | Apr 8, 2025 |
Base Model | |
Trigger Words | 3dstylev4 3d |
Hash | AutoV2 20930BD217 |
Yet another LORA trying to make a 3D style for Illustrious.
I recommend using ADetailer/hi-res fix
For some characters if the result looks 2D I would recommend lowering the strength of it to 0.7 like this "(son_goku:0.7)", same with character loras of certain anime and cartoon characters, lower their strength or they might come out looking too 2D
UPDATE 2: made v4. I didn't necessarily make it with the intention of it being better than V3 but just to get a different style, if it does look any better that is up to you to decide, if V3 was good enough for your uses then I recommend you keep using V3
UPDATE: made V3 trained on a more balanced dataset, felt like V2 and V1 specially leaned too heavily towards one particular artist's style, so I made sure to balance the number of images per artists on this dataset. Trained on 250 images (50 images per 5 different artists). Not an expert at all, barely know what I'm doing, probably the last update I'm making for this LoRA.