v1.6 is a Itercomp add difference merge using sd_mecha, basically model A is illuNextv1.5, model B is itercomp and model C is sdxl base. Then i add CLIP back from illunextv1.5 to illunextv1.6.
Not sure how well diversed the model is or if it actually is as "powerful" as Noobai V1, i might have to go back and start over from noobai-v1 base in the next release.
For now, this has clearer and better details, some noise can be visible depending on your sampler, looking into how to fix that