New Image Generation Model - Early Access

This is an image generation model based on training from Illustrious-xl.

It utilizes the latest full Danbooru and e621 datasets for training, with native tags caption.

Note: The model name and other details are subject to change.

The stable test version can be downloaded from Civitai, with regular updates. The current release is the version that continues training for 55,000,000 steps. https://civitai.com/models/833294

This model is still undergoing training!!!

Important Information

This model is in training and uploaded without testing. It's synced directly with our internal development.
Users access at their own risk. No guarantees or support provided.
No sublicensing or redistribution allowed. Approved users only.
We reserve the right to modify terms or terminate access anytime.
Use at your own discretion and responsibility.

Current Status

This is an early test version intended for internal use. However, we are considering allowing limited external testing.

Datasets

Latest Danbooru images up to the training date
E621 images e621-2024-webp-4Mpixel dataset on Hugging Face

Caption

<1girl/1boy/1other/...>, <character>, <series>, <artists>, <special tags>, <general tags>

Quality Tags

For quality tags, we evaluated image popularity through the following process:

Data normalization based on various sources and ratings.
Application of time-based decay coefficients according to date recency.
Ranking of images within the entire dataset based on this processing.

Our ultimate goal is to ensure that quality tags effectively track user preferences in recent years.

Percentile Range	Quality Tags
> 95th	masterpiece
> 85th, <= 95th	best quality
> 60th, <= 85th	good quality
> 30th, <= 60th	normal quality
<= 30th	worst quality

How to Apply

If you're interested in early access testing:

Please contact us through our official channels (e.g., QQ group).
Applications will be reviewed on a case-by-case basis.
Communication qq groups：①875042008 ②914818692 ③635772191