These loras were trained on images from Megami Magazine from late 2018 to early 2024.
V3 Type A: Trained on 50 images selected for their similar eye style and detailed backgrounds. The lora is great at prompt comprehension, amazing backgrounds and decent hands and feet. I trained with a few images from V2 and some others from the 1200 dataset I prepared. More than anything, the style is different from V2 and V1, so use the version you like best. I call it "Type A" because I'm considering training different styles using different images from dataset. Trained on AutismMix.
V2: This lora was trained on 150+ images that I painstakingly removed the japanese text from. (V1 is plagued with this, V2 and V3 had the text removed from the images) This version does well with characters and backgrounds. Trained on AutismMix.
V1: Trained on 1,200+ images. (All I had.) Trained on Based Pony.
To get the results you see in the images above, these are the settings I used.
Prompt(at the start): score_9,score_8_up,score_7_up,score_6_up,source_anime, <lora:MeMaXL_V3:1> [your prompt]
Negatives: mosaic censoring,bar censor,censored,source_cartoon,source_pony,source_furry,simple background,
Sample: DPM++ 2S a
Steps: 26 (Whatever gives you the best results)
CFG Scale: 7
Extension: [sd-webui-color-enhance] set to 0.5 (this increases the saturation a bit)
Upscaling: Follow this method
I exclusively use AutismMix and I recommend you do the same. All of the examples were made with that model.