[SussyBaka XL project has been shut down in response to NoobAI's licensing issue. The project is now replaced with AkashicPulse, which is based on RouWei, with better generation results and better license]
SussyBaka XL (SBXL) is an experimental fine-tune of NoobAI XL v1.0, codenamed Project PAI.
v1.0: Based on NoobAI Epsilon v1.1 Cyberfix with 10 extra training epoch, plus an extra fixing procedure, which made the model superior compared to NoobAI v1.1.
Steps: 25-30
CFG Scale: 4.0
Euler a
===
v0.2: This model shows significant improvement over the previous model, now outputs are more stable and almost no bad outputs.
===
v0.1: This model shows mixed results - while it can produce superior outputs compared to its base in certain scenarios, it occasionally generates significantly worse results, particularly with plain white backgrounds. Optimal settings:
Steps: 20-29 (recommended)
CFG Scale: 4.0
Known issues:
Inconsistent quality on white/simple backgrounds
Performance varies greatly between generations
Released under Fair AI Public License.
Note: This is an experimental model, and future improvements are planned pending funding availability.
Description
Based on NoobAI Epsilon v1.1 Cyberfix with 10 extra training epoch, plus an extra fixing procedure, which made the model superior compared to NoobAI v1.1.
FAQ
Comments (2)
"10 extra training epoch"
what amount of images per epoch? What is dataset?
What amount of images per epoch? ~3.500 images. It was a Unet fine-tuning (text encoder off).
Dataset? A set of manually curated images I collected from selected art style.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.





