28/10/2026 -V2
V2 Improves the overall look by:

Makes it more amateur photo style (but can be taken further with loras)
Lightens the default skin tone

Closer shots by default

Reduced grain.

Improves eyes.
Recommended 12 steps, no extra speed LoRA needed.

24/10/2026 - I uploaded a new version of int8 as the old version had learned_rounding turned on (to improve quality) that made it nearly 4 times slower at inference.

Also now fixed my workflows , as they had been exported in API mode.

Makes realistic, sharper images with Moss Image
It has the 8 step Lora merged in , I run it at 4- 9 steps er_sde/Bong_tangent, cfg =1

-can ignore other artistic styles in the prompt.

I think it is a banging good model :)

Check out my Basic and Advanced workflows >