V1 is trained on ±1200 images. It does require Hi-res fix to look decent. ADetailer is optional but gives even better results. The model should work without a trigger word, but works better when you do use it. I have only tested it on more realistic checkpoints, so I do not know what this LORA will look like when used with a 2D Checkpoint.
V1 is trained on ±1200 images. It does require Hi-res fix to look decent. ADetailer is optional but gives even better results. The model should work without a trigger word, but works better when you do use it. I have only tested it on more realistic checkpoints, so I do not know what this LORA will look like when used with a 2D Checkpoint.