
FLUX.ORBIT ALPHA TEST
Experimental LoRAs to add:
-> Good furry aesthetic
-> NSFW concepts (very unstable for now)
recommended mix:
<lora:fluxorbit_large_V4_000015000:0.8> <lora:fluxorbit_medium_nsfw_v1_000005400:0.2> <lora:fluxorbit_mini_v2:0.3>
recommended settings:
Sampling method:
- SFW: [Forge] Flux Realistic
- NSFW: Euler
Sampling steps: 20
Width: 720
Height: 920
Distilled CFG Scale: 3
CFG Scale: 2
Warning:
-
The model is very biased towards anthro foxes, muscular characters and big spiky hair
-
The NSFW part is very unstable for now, you might need to generate like 100 pictures to get 5 good ones
-
most stable is front view
-
back and full side views are almost impossible to do
-
How to prompt it ?
-> It's a blend between Pony and Flux
Everything is all over the place and weird, what should I do ?
-> You can try ranking up the CFG scale up to 3 or 3.5. If it's still not enough, you can go up to 7and put those settings in DynamicThresholding (CFG-Fix) Integrated :
描述:
Works well but very stiff and biased
训练词语: fluxorbit
名称: fluxorbit_mini_v2.safetensors
大小 (KB): 335748
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success