About

I proudly introduce you to XiXi, which is the second model contribution to Civitai. Where my previous model Ka-Chu was trained on WAI-REAL_CN V6, this one is trained on its older version WAI-REAL_CN V4. I made this conclusion after a discussion with ruanyi, an author who makes exemplary Pony content. The next V2 version of XiXi, might be trained on WAI-REAL_CN V6 or up or PDXL.

Training

XiXi V1 is trained with 28 images (square 1024px) with (8 repeats, 2 batch) on Civitai with a max epoch of 21. Since her source faces are expressionless and have a minimal expression, I concluded that it was starting to look nice at epoch 7 already, but in the end I ended up with epoch 15.

Recommended Setting

• Steps: 24-28

• CFG scale: 6.5-7

• Sampler: Euler a

• ADetailer face_yolov8n/s.pt

Adetailer usage with other Lora's

Copy paste the whole prompt in Adetailer when using face enhancement, and remove all Lora prompts that does or could do something with faces and keep this one.

Works with other Lora's

I am a big fan of Ruanyi's work. Most of Ruanyi's models can be used with XiXi, but imaginations to you.

If you like my work, don't hesitate to leave a like or a review. Please share your renders so I gain knowledge what other potentials this model can be used for.

Last but not least, I drink a lot of Ko-Fi :)