This is a blend of about 9 different models using a combination of direct merges, block weighted merges, and some Loras integration. The focus for this model was faithful reproduction of human faces and backgrounds. It responds we to most prompts and typically does not require the use of face restoration. Negative embeddings such as Easy Negative or Beyond Negative may improve performance, but can reduce the model's ability to properly accept Loras. The model is reasonably receptive to most Loras, but handles ones intended for realism best. For best results, total Lora weight should not exceed 1.4, although it has more tolerance for detail improvement or sharpening Loras.