[v3.1 is still in further testing. Updates regarding new findings will be updated in the "About this version" section]
UrangDiffusion v3.1 (oo-raw-ng Diffusion) is the first UrangDiffusion version that utilize Animagine XL 4.0 as the base.
The name “Urang” comes from Sundanese, meaning “We/Our/I.” The history behind the name is to make the model not only suitable for me but also for many people. Another reason is that I use many resources (training scripts, dataset collecting scripts, etc.) from other people. It’s unfair to claim this model as “my sole work.”
Standard Prompting Guidelines
Prompting guide:
Default negative prompt:
lowres, bad anatomy, bad hands, text, error, missing finger, extra digits, fewer digits, cropped, worst quality, low quality, low score, bad score, average score, signature, watermark, username, blurryDefault configuration: Euler a with around 25-30 steps, CFG 5-7, and ENSD set to 31337. Sweet spot is around 28 steps and CFG 6.
Training Configurations
Finetuned from: Animagine XL 4.0 Base (NOT 4.0-Zero)
Finetuning:
Dataset size: ~1,600 images
GPU: 1xA100 80GB
Optimizer: AdaFactor
Unet Learning Rate: 1.25e-6
Text Encoder Learning Rate: N/A (Turned off)
Batch Size: 48
Gradient Accumulation: 1
Warmup steps: 5%
Min SNR: 5
Epoch: 15
FAQ
Q: Images are sometimes noisy.
A: This is a common issue with Animagine XL 4.0 models in general. The base model is trained with only 10 epochs, which lead to the model being undertrained. Unlike Initial N or Initial I model that are trained with more resources.
Q: Hires fix model?
A: Check out the cover image metadata, you'll find it there.
Q: Initial N/Initial I is better.
A: Just leave and do not use the model. Simple. No need to announce your departure. Except you're willing to leave a constructive feedback or willing to fund future projects.
Special Thanks
My co-workers(?) at CagliostroLab for the insights and feedback.
Nur Hikari and Vanilla Latte for quality control.
Linaqruf, my tutor and role model in AI-generated images, and also the person behind tag ordering.
License
UrangDiffusion v1.0-v2.5 falls under the Fair AI Public License 1.0-SD license, while v3.x falls under the CreativeML OpenRAIL++-M license.
Description
[TL;DR: New finetune dataset]
Dataset size: 2,000 images
GPU: 1xA100
Optimizer: AdaFactor
Unet Learning Rate: 1.75e-6
Text Encoder Learning Rate: - (Train TE set to False)
Batch Size: 48
Gradient Accumulation: 1
Epoch: 10
Noise Offset: 0.0357
FAQ
Comments (5)
Hi, in my tests the model performed well for generating art in a style that suits me. Could you tell me what goal you want to achieve with this model? Greater knowledge of characters, maybe a stable beautiful style, or perhaps training it to understand more concepts and create more complex scenes. From the description, I only saw that you added characters from games and anime.
Hi, for the main goal of the model is to finetune the model to have it generate images in my preferred style and trying to make it generate better scenery , since the original Animagine XL didn't really fit my likings. As for the character addition, there are two reasons for adding them:
1. As my model training practice, since I'm working on something else much bigger than this. (Spoiler: I work for CagliostroLab)
2. I enjoy playing Wuthering Waves and Zenless Zone Zero in my spare time, and watching hololive. So I wanna make sure my model can generate the characters I like 😅
@kayfahaarukku Thank you for the response. I'll take a closer look at what the model is capable of in the coming days, and if I notice any issues, I'll share them.
Regarding working on something bigger, I'm on the CagliostroLab Discord server. Recently, they were asking which characters from which titles users would like to see and were gathering a dataset. It's clear that this will either be for Animagine 4 or a new model, and it's not far off. I'll see what happens.
good work! been following the aing models for a long time and i cant wait to see how this sequel evolves
I'm no expert, and I only had some limited time to play with the model, but so far, I'm greatly impressed! I love how it has a definite stable style and when you use artist names, it responds amazingly! Mixing multiple artists as well!