A merged model built from multiple LoRAs with carefully balanced weights. After extensive testing across different runs, the model demonstrates strong stability and consistently high-quality results.
The full workflow is provided within the preview image.
Recommended samplers:
DPM++ 2M SDE GPU (SGM Uniform)
Euler A (SGM Uniform)
LCM (SGM Uniform)
Feel free to experiment with other samplers as well.
This is the initial release (v1), with potential improvements and future versions planned.
NSFW content supported.
Description
FAQ
Comments (6)
Hi, can you provide a BF16 version to make other quant like mxfp8?
Do you mean of the version I publish? this version is backed from initial fp8 variant so no bf16.
@png oh sad D":
"built from multiple LoRAs"
which loras? lol
The bigs ones :)
Very well done, keep it up with the good work. ♥