A NEW BEGINNING
Starting with a new Base from curated Training Set. It is capable of almost everything and not polished in any direction yet. It should have good prompt adherence and diversity with new concept outputs.
Upcoming for V2
I plan to introduce some classification to the next Version for the styles "abstract, cinematic, drawing, illustration, painting, photo, render, vector_art" which are hand curated on the Dataset of ehristoforu
Also planned to adding more details to the generations
⭐ Feel free to leave Reviews and Samples - and always have fun creating ❤
Description
FAQ
Comments (15)
Great Idea! Love it ;)
epinikion, thank you so much for everything you do!
I know epic checkpoints stand for quality, but as this is only v1 i wasnt expecting too much yet. But it really surprised me how good the quality is. Really outstanding!
Some great details I find the eyes are off on most full or 3/4 body subjects, close ups seem to do well. Figuring best way to use vs your other XL models (that I really like) but could be settings, any pref to the setting or difference in how it should be prompted? Might just be seeds of my usual test images as well so will try more out. :)
Lower cfg see,s to really help for me the 4 to 5.5 range if subject is not a closeup. Great results, makes it hard to decide :)
The old epicrealism and other models, like dreamshaper, Juggernaut achieve realism by contaminating the pixels. We need smoother skin texture, while keeping realism.
And, it would be better to release a hyper low step version with each release, for those with poor GPU.
Very cool, looking forward to it!
So Far So Good, face ip adapter work well with this CP..
great job
Since the community seems only interested in Flux cuz of extremely easy Lora making, and also on Pony.
Any chance we can see real Fine Tuning of SD 3.5 by Epicrealism?
I'd much rather see a real flux finetune. SD3.5 appears to be dead on arrival.
@Polygon All the flux finetunes that are being uploaded at present are basically people merging their Loras into it. Anyway, I did have success in making good Loras on SD 3.5, but seems very less demand.
Looks like I probably should have made that suburb liminal lora on this one. So I will!
Try it out, but i still think, training on base SDXL Model always works best
still waiting for the next version..
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.