• Version 1.1 uses 100k Pony CLIP Models and is PONY lora focused.

• EA/Beta uses a custom Laion training of CLIP-G and is SDXL lora focused

• 832x1216 & 1280x1536 native generation size even 1280x1792 is possible for highly trained subjects

• Hybrid Model version is the same as BF16 but with FP32 CLIP and FP32 VAE (Auto1111 has an issue with stored BF16 VAE)

• The best quality will come from forcing unet in BF16 not FP16 COMFY --bf16-unet - Auto1111/Forge --unet-in-bf16

This model took over 120KwH of energy to train. Not including the 18KwH used to train the CLIP models.

Steps: 20+

CFG scale: 5

This is not my model, i take no credit for it. all credit goes to the creator @ https://civitai.com/models/1615041?modelVersionId=1971351