Consistent Factor (Euclid) Reviews
Easily the most photorealistic and LORA friendly model I've used. With proper training I don't need any additional textual inversions to create quality images. Quite literally can use trigger word and one or two other tags I trained and get exactly what I'm looking for. I've tried other models but can't find anything as... consistent... as this one. Goated.
Model is very good at realistic faces and body poses. It is possible to get feet with 5 toes and hands with reasonable number of fingers. And as a plus it has a tendency to get this dark gritty look in pictures, so not everybody looks like they are made of plastic.
The merging/training definitely has biased the model to prefer to generate human faces (and often very similar looking ones). When I tried reusing some of my prompts that with other sd1.5 models generate cute monsters and animals by default, it tend to generate humans in cosplay costumes with this model instead. As some of my examples show it is still possible to get alien monsters and hybrids with this model, it just takes a bit of extra prompting compared to other models.
On a positive note the lighting and photo-realism is great. It can go pretty dark and gritty in the output something that sd1.5 models often struggle with so good job on that aspect.
