Eleusis

Hundreds of years before Christ, political leaders, philosophers and ordinary folk made a once-in-a-lifetime pilgrimage to the Temple of Eleusis to partake of what was purported to be a life-changing visionary elixir whose ingredients are debated even today. Like that magical vision-inducing drink, this model is a mixture of some of the finest image-generating “ingredients” on offer today, whose contributions are gratefully acknowledged.

Eleusis v1.0 is a merge of an early version of my custom eye LoRA, with varying amounts of BastardsV1, nightvisionXLPhotorealisticPortrait_v0791Bakedvae, and JuggernautXL_9Rdphoto2Lightning. Consequently, it does well with low step count (around 8) and cfg values of 2.5 – 3.0.

It has been tested with several samplers and seems to perform well with Euler_ancestral, dpmpp_sde, dpmpp_sde_gpu, dpmpp_2m and dpmpp_2msde_gpu. I prefer karras as a scheduler, but encourage you to experiment! Have fun, post your results (absolutely no NSFW please!)

You will find Eleusis v1.0 is capable of generating a broad range of images of interest to the AI community. Male and female portraiture, flowers, birds, and concept art are all there for the asking.

By using this model you are tacitly, if not explicitly agreeing to be bound by the licenses and terms of use set by the authors of the component models.

A HUGE thank you goes out to Triple Headed Monkey, without whose help and encouragement, this model would not ever have reached the AI community. Additional shout-outs and thanks to speach1sdef178, Grokster and Pauluz RTX 3080ti for investing their time, effort and resources in testing and evaluating the model pre-publication. Last, but not least - to Afroman4Peace - who is an inspiration to all who aspire to advance the creativity of the AI community. They all deserve your support, which I hope will be given freely in exchange for their generosity and vision.