SDXL LoRA/LyCORIS works best on its original training model. You may need to reduce weight when using it with other models or additional LoRAs.
LyCORIS Combination Recommendation: Try using four of my LyCORIS models with the following weights: 0.4, 0.3, 0.2, 0.2 (or any combination with a total weight around 1.0).
I will release multiple versions for a few popular models. Feel free to request for artist or model.
Version naming convention: MajorVersion.MinorVersion [BaseModel]
MajorVesion: For big updates that can apply to any style LoRA. For example, training parameter update.
MinorVersion: For small updates that only apply to this LoRA. For example, epoch selection, removing a certain image from dataset.
BaseModel: SDXL LoRA/LyCORIS works best on the model that it was trained on.
Description
FAQ
Comments (11)
Could you make a Yang-Do (OrangeMaru) style lora for AutismMix? It's another model based on PonyXL just like AuroraXL. From what I've seen there are 3 different versions of that model and apparently AutismMixPony is the best for lora training.
@GogetaSSGSS3 Thanks for your suggestions. I will test training on AutismMix. I will start with artists that I have trained previously for comparison.
Here are the test results: https://civitai.com/posts/1460580
I trained the same artist on four different base models: auroraXL, JS2Pony, ponydiffusionV6, AutismMixPony
@silvermoong Interesting, from what I can see it seems like Aurora looks the best from that list. The main issue that I have with models such as Pony/Autism is their base models having classic AOM style look (at least that's the vibe they give me). That could be because it was trained on a lot of western style images. But overall I think all of these Pony models seem to be good when it comes to understanding loras. All of these models have their strengths and weaknesses I guess. Even tho these models know a lot of nsfw concepts/characters their main weakness is the artstyle, so I hope you continue to make more of these style loras :)
@GogetaSSGSS3 I have been testing AuroraXL and JS2Prony with more artists. I found an issue on AuroraXL and am currently checking if JS2Prony could be the bettter one.
Please train gemba on Pony diffusion or AutismMixPony
Hi, did you mean https://www.pixiv.net/users/9965434?
I'm not sure how well it a style LoRA could learn poses, I will try later.
I'm comparing a few base models for LoRA-training. Here are some results: https://civitai.com/posts/1460580
Personally, I feel aurora XL learns artist styles best.
@silvermoong aurora is based on pony
@Rating_Agent Yes. All models in the test are based on pony. I was looking for the best pony based model for style LoRA training.