Style Lora in the Style of AHy2. Request. I dont know the Artist. The datasets look like it was already Ai generated. Works very good with other lora but on its own it likes to gen Hair ornaments and Hairstyle based on hair color. Like to hear other opinons, so i can try to train an other version.
Added v2.0
Trained on illustriousXL_v2.0
Less stiff in poses
Works better overall
V3.0 Test and v4.0 Test :
Both trained with 2000 (older) extra images
v3.0 is with 1 trigger word and v4.0 like v1 and v2
in my opinion it doesnt make the "Style" better and its just a generic Anime style now.
Since the "Artist" made anything with AI only the newer imaes had something "Special" for me
But i trained it so u can try if u want
Illustrious :
v1 :
Trained on illustriousXL_v01
Weight 1.2 - 1
Trigger Female : 1girl, large breasts, looking at viewer, solo focus, blush, bangs
Neg Prompt : "crown, hair ornament"
v2 :
Trained on illustriousXL_v2.0
Weight 1.2 - 1
No trigger
Neg Prompt : "crown, hair ornament"
V3 Test and v4 Test :
Trained on illustriousXL_v2.0
Weight 1.2 - 1
V3 Trigger : ahy2
Would be awesome if u leave a like post a pic ;)
Description
FAQ
Comments (9)
Wow it works great, thank you so much 🥰
我感觉这位ai作者在画风中融入了些许Piromizu作品的画风
All dataset images are from the same autor. if it looks wrong, im sry.
Lora's is already good, but the color is not very thick and her face is not detailed enough, try checking AHY2's pixiv to see more details, that's my suggestion 😁
I left pixiv cuz of too much pedos. i collected 444 images. perhaps older and newer stuff. so in the end its a combination of all. i dont think the faces in the dataset images are so much different. the shading and lightning are good. what kind of background and colors u use is your choice. i think there are some tags that enhance contrast and colors. and i trained it in illu v2 but made the images in an other checkpoint. so that waters the result a bit too. im ok with it ;)
@BloodySunkist Well, even though there are a lot of good arts on pixiv, in my opinion, Ahy and Ahy2's art styles are different. Ahy's colors are flat and Ahy2's colors are more colorful and bright, so if they are mixed, they don't match. How can Ahy make so many characters with the same art style? Does he use special AI? I'm still a beginner in the AI field 😁, and I used WAI-NSFW-illustrious-SDXL for the model
@crosetroser45141 I used all images from ahy2 as far as i know. i didnt know ther is a ahy1 :D. i think most artist paint digitally so keeping a style constant is no problem. and im sure he/she can paint really good with hand so im happy there are still lot of talented people out there. without them ai would be even more worthless. yeah a lot of people use WAI-NSFW-illustrious-SDXL. i dont know why, didnt think its special. or i just like more underrated checkpoints. but mostly style lore look closesest to the style in the checkpoint they are trained in. so mostly illustrious basemodel
@BloodySunkist AHY's account has been deleted and then changed to AHY2, AHY's art style is flat while AHY2's latest art style is more colorful, like vivid color. Apart from Civitai, I also use Seaart, but your Lora is not on Seaart. Don't you play Seaart?
@crosetroser45141 seaart is the only other site i have uploads. but only the one other have stolen and i could claim. since nobody published this lora there i uploaded it for u. but its a fucking pain in the ass to upload on that shity site. https://www.seaart.ai/de/models/detail/d2ovj85e878c73962bug