A model of Miniami Fuyuki from the anime Dosanko Gal wa Namara Menkoi. Its my first model that I'm uploading to civitai, hope you guys enjoy it. Please let me know if you want me to keep making models by posting some pictures you make with this Lora.
Since everyone was asking me to make the models in Pony, i also made a Pony version. Let me know if you like it and want me to keep making Pony models.
Use strength of 1.0 for pony model.
Strength used: 0.7, can go up to 0.8
Use the below words to get the different outfits:
outfit 1 : cleavage,school uniform,scarf,red scarf,blue jacket,open clothes,long sleeves,neck ribbon,pink cardigan,shirt,grey skirt
outfit 2 : cleavage,green_ jacket , clothing_cutout, cleavage_cutout , turtleneck, blue_sweater
outfit 3 : cleavage, jewelry, collarbone, jacket, earrings, track_pants, heart_necklace, hood, black_jacket, hoodie, black_pants
outfit 4: hair_ornament, cleavage, bare_shoulders, underwear, ponytail, heart_necklace, pink_bra, pink_shorts, pink camisole
Use WITH Adetailer and WITHOUT face restoration.
Enjoy without moderation.
Description
FAQ
Comments (8)
May I ask do you have any tips for SDXL (pony) training?
I learned from this article : https://civitai.com/articles/138/making-a-lora-is-like-baking-a-cake
To train Pony XL I use the pony XL V6 checkpoint , together with this Konya config : https://files.catbox.moe/ntdmiq.json.
I changed 3 things in the Konya config : Network dimension to 16 instead of 64 and network alpha to 8 from 1 and epochs from 5 to 20. i usually use 300 steps per character ( for example if you have a database of 100 character pictures, i let them train 3 times on each picture, so 300 steps in total per epoch).
It is a lot of work to extract all the character screenshots from the anime, so bear in mind that creating loras is a very long process. Also i m training on an RTX4090 which takes about 1hour of training per character AFTER all the screenshot taking and captioning.
Hope this helps!
Thank you for the tip. I'm wondering if you have any additional tips on how to extract character screenshots from anime more quickly?
@JB2023 I do everything frame by frame, and also always try to take the frames where there are no subtitles, so basically its a lot of work. There might be some programs that do it automatically, but i haven't found any yet.
@ZedoTasco ok got it. I think your models are great. Hope you can release more characters in the future!
@JB2023 Thanks for your feedback. I for sure will!
Hey there, may I ask which UI you use/recommend for SDXL? Is this all a1111 or comfyUI? I tried recreating your images with a1111 but they're not quite the same as the samples
i m using stock A111 with Adetailer on, nearest upscaler 2X and R-ESRGAN 4x+ Anime6B at 0.5. If you are running on a 8VRAM card it might be worth checking out the Forge Webui.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.