Illustrious

Weaker Style LoRa, requires usage of more style based keywords such as anime_coloring,anime_keyvisual and anime_screencap. Overall quality is better than animagine but style application is weak. Most of the changes are rather subtle such as color palette preferences. This LoRa mostly has control over the color palette rather than shading. LoRActl at [email protected],[email protected] steps should be good enough to prevent most concept bleeding while retaining the style.

Animagine

Trained on 205 Images at 1024x1024 and 1024x1536 resolutions with AnimagineXL3.1. The images were generated from AnyLoRA merged together with different artist LoRAs. The result is very similar to BlueAilandMix so I never released it (Hands are pretty bad with merge). I seem to be understanding SDXL training better than merging so I made a LoRA instead. This is my first time making a style LoRA so I am expecting issues with it.

My aim was to create a simplistic flat-shaded artstyle with minimum detail around the hair and better lineart quality since I wasn't satisfied with animagine's default anime_screencap.

Trigger word: Ailand (This LoRA is fairly weak so it is highly recommended that ailand is first word in the prompt!)

Recommended Weight:1

While animagine quality keywords are still usable, I don't recommend using them since it may suppress the effects of the LoRa. anime_coloring, and anime_keyvisual seem to work fairly well together with this LoRA. I don't recommend using anime_screencap.

Problems:

• Pixelization of the lineart (Requires a hi-res fix of 2x or inpainting pass to fix)

• Bad Anatomy

• Bad Hands