[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
Training Configurations
Finetuned from: Animagine XL 3.1
Pretraining:
Dataset size: ~23,500 images
GPU: 1xA100 80GB
Optimizer: AdaFactor
Unet Learning Rate: 3.75e-6
Text Encoder Learning Rate: 1.875e-6
Batch Size: 48
Gradient Accumulation: 1
Warmup steps: 100 steps
Min SNR: 5
Epoch: 10 (epoch 9 is used)
Finetuning:
Dataset size: ~6,800 images
GPU: 1xA100 80GB
Optimizer: AdaFactor
Unet Learning Rate: 2e-6
Text Encoder Learning Rate: - (Train TE set to False)
Batch Size: 48
Gradient Accumulation: 1
Warmup steps: 5%
Min SNR: 5
Epoch: 10
Noise Offset: 0.0357
Wuthering Waves, Zenless Zone Zero and hololiveEN -Justice- have been added to the model.
FAQ
Comments (6)
Is there a list of characters added to this model?
Wuthering Waves, Zenless Zone Zero and hololiveEN -Justice-.
I don't have an exact list, but I updated the dataset about 2-3 weeks ago. For Wuthering Waves, it should include data up to the Changli banner. For Zenless Zone Zero, it includes information up to the Ellen Joe banner. And all 4 members of hololiveEN -Justice-: Elizabeth R. B, Gigi Murin, Cecilia Immergreen, and Raora Panthera.
@kayfahaarukku Thanks for the info, I tired to recreate Nicole, Lucy and some characters from WuWa, but with minimal results, Changli worked easily, but Ellen is a hit or miss. So that is why I was wondering if they had specific triggers, but I will play around more. Thanks for the Model BTW!
@darionk Thank you for the feedback. Nicole should be the easiest to generate because the training dataset was around 450 images; use "nicole demara, zenless zone zero" in your prompt.
Lucy, on the other hand, has a training dataset of less than 100 images when the dataset was collected. Unfortunately, it will not work without LoRA.
Ellen's training dataset size was barely reaching the generation minimum, so it will be hit or miss.
The same thing is happening with the WuWa dataset. The number of images trained was barely reaching the generation minimum.
I will try to fix this in the future by updating the dataset. However, I can't promise that it will fix everything.
@kayfahaarukku That is fine, thank you for the hard work, I was just wondering why some characters weren't working. Nicole makes sense since I haven't used Demara, I should try that. I wonder who else has that, I will keep playing around. Thanks for the fast responses, I appreciate the info!
@darionk You can check danbooru if you're not sure what is the full tag of a certain character name. Keep trying and cheers!