This LoRA model has been trained on Kohya SS based on Cherry Mouse Street artstyle on WAI-NSFW-illustrious-SDXL checkpoint. Obtained graphics could be really close the original art style. This LoRA model could be use for cartoon representation of sexy women.
This model was my first LoRAS training test using an Illustrious compatible checkpoint, and actually the result is really positive. So, from now, I will start to convert my old Pony LoRA model for Illustrious based on their rate/popularity.
Note : While the training dataset includes works created by Cherry Mouse Street without direct authorization, this LoRA model's design ensures full compliance with copyright laws. This is achieved through a transformative training process that abstracts and generalizes Cherry Mouse Street artistic style rather than replicating specific copyrighted works. As a result, this LoRA model is able to generate innovative and original content inspired by Cherry Mouse Street artistic influences, while remaining within the boundaries of legal and ethical standards.
Description
FAQ
Comments (10)
Looking very good with your first illustrious lora, I did hear opinions that illustrious is better at recreating artist style than pony. What do you think?
Actually I have just trainged tihs one. I could give you a more relevant opinion once I have trained around ten models. What I can tell you actually is that Illustrious LoRA artstyles seems a little better than Pony ones, but require more training steps (+50%).
@GTSuya_Studio i see, so i am guessing its a bit more expensive and time consuming to train pony lora i guess, i mean for the better quality i do think its a fair trade off
@Sonziggy Some update of my first impression. Training non-amine artstyle with Illustrious checkpoints seems more difficult than with Pony. I suspect that Illustrious checkpoints have been intensively trained on manga/anime ressources (like Dandooru), and that it is difficult to train western artstyle with them. See for example my Paolo Eleuteri Serpieri artstyle for Illustrious (https://civitai.com/models/1101031) anmd compare it with my Pony one (https://civitai.com/models/710508). Pony seems more flexible with non-manga/anime artstyles.
@GTSuya_Studio i mean that makes sense to me, even i encountered the same issue when trying to get artist styles just through prompt, its usually the eyes, they appear anime-ish but its not that bad because danbooru also has a lot of non-anime/western styles and i did have a lot of success with artists like "nuclear wasabi" just through prompt so getting the style through lora would be even more effective. The problem is just the sheer amout of anime styles on danbooru which gives illust checkpoints some bias towards anime like styles, i think using negative prompts might help?
@Sonziggy I have tried with negative prompt and it did not help. As it is mainly trained with Danbooru tag, there is no "Anime" tag to put into the negative prompt tov force non-anime/western styles. I have also try to put western_comics_(style) into positive prompt, but it did not drasticaly improve results. So my conclusion is that Illustrious checkpoints are just good for anime/mange artstyle. If you want to use non-anime/western artstyle, it is better to use Pony.
Found this on Dreamerland application. Can this be used for commercial purposes?
This LoRA file is based on Cherry Mouse Street artworks. On our side it is okay for us is you use it for commercial use (it is indicated at this on Civitai). BUT, we did not ask Cherry Mouse Street for permission to use his artworks for this Lora model training. So, if you plain to use it for commercial purpose, you may have a problem not with us, but with Cherry Mouse Street himself.
Best move would be to ask directly to Cherry Mouse Street. Some artist are complacent with the use of their works by AI (they see it like a free promotion or their works). But others are not at all.
Awesome style!
I'm trying to recreate a similar style as the picture in the preview in stable diffusion. Are there other loras I need to use?
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.