Umamusume All in One LoRA
This is an All-in-One LoRA designed to generate a wide variety of characters from the popular game Umamusume: Pretty Derby with a single model. Without the hassle of managing individual character LoRAs, you can easily bring numerous Umamusume to life, including Stay Gold, Almond Eye, Loves Only You, and many more, with just this one file.
Notice
This model is a re-upload. The previous model page was accidentally deleted during a version update process. My apologies for the inconvenience!
✨ Features
Extensive Character Library: Capable of generating most of the main characters from the game with high fidelity.
Versatile Costume Support: Faithfully reproduces each character's unique racing outfits, as well as school uniforms, casual wear, and other styles.
Easy to Use: Achieve your desired images simply by inputting character names or feature tags, without the need for complex prompts.
🌟 Character & Prompt Information
For detailed prompt information regarding the characters, costumes, and features included in this LoRA, please refer to the Hugging Face link. Find the precise tags for the character you wish to create and utilize them in your prompts.
License
This model is released under "Derivative work guidelines for umamusume".
And "Fair-AI public license 1.0-SD" License
Description
This is pony-uaf version of SDXL AIO LoRA of Uma Musume
Recommended options
LoRA weight 1.0
Trigger words
[character name] \(umamusume\) or [character name]
For example,
manhattan cafe \(umamusume\)
hayakawa tazuna
Settings
For Stable Diffusion XL Pony V6
Please follow pony's generation guide.
SDXL-Pony version
trained on sd-scripts by kohya_ss and LyCORIS by KohakuBlueleaf
Thank you all a lot!
Base model : Pony Diffusion V6 XL
Hardware : 4x A6000 48GB
Training time : 39 hours
Dataset : ULTIMA Dataset + knowledge cutoff: 2024-04 (uploading is soon)
With synthetic natural language caption. half-half ratio
Max Token Length : 225
Resolution : 1024x1024
Dim, Alpha : 32
Algorithm : LoCon
Network dropout = 0.1
Optimizer : Lion8bit
weight decay=0.1, betas=(0.9, 0.95)
Steps : 10,777 (13 epochs)
Total Batch size : 64
4 GPU x 8 batch size x 2 Gradient Accumulation steps
Learning rate :
U-Net : 8e-5
Text Encoder 1 : 4e-5
Text Encoder 2 : 4e-5
warmup to 1e-4 for 100 steps and then kept constantLR scheduler
LR scheduler : constant with warmup.
Min SNR gamma : 5
Noise Offset : 0.0357
FAQ
Comments (33)
Yakuza are on ur way buddy
Glad to see your new works. Will you release SD1.5 version separately like your previous works?
If resource allows, I will try updating 1.5.
seems the AIO one for SD1.5 could not work well as those separate ones.....
what model did you try?
@mht those lora and loha you made, such as for mcqueen and rickey
@mht especially used with other lora together
@OathOfJudah you should use lora or loha only one at a time. if you combine two lora, unexpected result come out. or control strength carefully.
And 1.5 version is updated
Is Pony_uaf not available with AnimagineXL?
It shouldn't use a weight of 0.6
Is there a separate AnimagineXL version?
-pony- version is for pony model. you should use sdxl-alpha version.
@mht I tried SDXL-alpha and AnimagineXL, but lora doesn't work, and I prefer AnimagineXL because it's a bit difficult to use pony, so I'm waiting for AnimagineXL-uaf.
what abou kohaku-series? it is similar with animagine and uses tag-based prompt
@mht I'll use the kohaku version and wait for the AnimagineXL-uaf version to come out, I saw in another comment that you're making an AnimagineXL-uaf version.
Uploaded!
@mht Awesome!! Thank you!! Love you!!!
@mhtÂ
Is it possible to update the ULTIMA-prompts?
I'd like to see more detail in the new character's jewelry and such.
Also, the color of the star earrings on Narita Top Road doesn't change no matter how much I put in, is there any way to change it other than inpainting?
SDXL-animagine-uaf Thank you so much for the update :)
prompt updated!
+) due to dataset bias, object bias of character is natural. Even I refine captions, model is working upon conditional probability. So model understands Narita Top Road always has those earrings.
@mht Thanks again for the quick update :)
I asked because the color of the starburst ornament is blue, not purple, and the 1.5D was good at prompting or announcing the color.
Thanks. I will put it to good use. love it!!!
Hello, I am a big fan of your lora and have been for a long time!
Are there any plans to create a UAF version that is compatible with AnimagineXL?
Yes I have. It takes a week.
@mht Yay! Thank you!
I am very much looking forward to it!
I will always be rooting for you.
Uploaded!
@mht thank you ! I will enjoy them from now on!
@mht This is best Lora and I am very impressed! Thank you so much!
By the way, how do I know which prompts call for costumes like UmaDiffusion/ultima-prompts for newly added characters?
@gotcha9999 I will update prompt doc for characters tomorrow.
@mht  Thanks! Looking forward to your updates!
updated!
@mht Thank you very much! It was very helpful! I will continue to support you!
Hi~ I am trying to use the online generator in civitai for Byerley Turk but turns out Turk is a prohibited prompt. Is there other way to trigger it?
sadly no way :(
@mht Ah...thank you for replying. I have sent a review to Civitai to see if they will make adjustment :3
Ah~I finally find the way to generate her with the online generator, Use Byerley Turk\(umamusume\) will be okay :3