A realistic merge with Illustrious and my SDXL merge models.
V4 Ascension
Better, more natural realism and lighting and skin detail.
This model outputs better with square images around 1280x1280.
V3 Rapture
This version has better skin realism and better faces.
Removes the washed-out colour saturation look happening on some images.
In my workflow I am using a CFG of 1.2 but on site it may need to be higher.
V2 Revelations
Better photo realism an works great with my tuned ComfyUI Workflow
V1 Genesis
It seems to like a low cfg of around 2-4 instead of the usual higher values.
Scheduler: Euler a/ancestral - Normal , 20+ steps. CFG 1.1 - 2.0
I would highly suggest you try my ComfyUI Workflow built for this model: https://civarchive.com/models/1255024?modelVersionId=1591700
Description
Initial Version
FAQ
Comments (8)
Amazed nobdy has commented yet so allow me! Great job. Thanks for the continued efforts and making this savagely hot realistic IL model, will watch with great interest. Pony has some serious competition now :D
Thanks, I am a bit late to the game with Illustrious models as I have bee focusing on Flux. But it is really great to see how far the SDXL architecture can be pushed. The outputs of this model are quite distinctive vs other realistic Illustrious or NSFW SDXL models, something quite in between.
Wow. I'm impressed. I like to generate realistic photo images of hot situations lowered in everyday life. With a certain light and faces. Thank you
Thank you, yes I am pretty happy with how this version came out, I think it is a bit different than any other model on Civitai at the moment. I do have a few ideas for pushing the photo realism and detail a bit further in the next release.
If you want to squeeze more realism from this model without using flux or other techniques as a refiner you can use DMD2 lora, ~16 steps, lcm, exponential, cfg 1.0. Pair Advanced Lying Sigma + Detail Daemon Sampler together. A little bit of Pertube Attention Guidance also helps. I'm also using illustriousFP32CLIP_L and CLIP_G.
Yeah thanks, I worked out basically that same workflow in a few hours of testing the other day (apart from CFG =1, I will try that later) , I also add a small amount of film grain with a node and tweak the saturation/contrast. I am also working on a version 2 which has better realism.
(((thoughts on my updated workflow for Jib Mix Illustrious Realistic)))
I ditched all the "heavy" upscalers (Supir, Ultimate SD, etc.) from my Illustrious workflow in favor of the "DMD2 Essential Upscaler/Refiner Workflow" https://civitai.com/models/1278122/dmd2-essential-upscalerrefiner-workflow. You can add a "Film Grain" node before the 4x upscale to introduce an extra layer of filmic realism.
Previously, I was using Jib's Flux model for the refinement process. Now, I’m using Illustrious, starting with the ~8-step DMD2, followed by a 1.25x upscale with the recommended (35-70) steps and ~0.25 denoise.
I removed all the IllustriousXLFP32 Clip L & G models from the workflow, as they don’t work well with the realistic Illustrious models.
Start using "Dynamic Thresholding" and "P.A.G" to see how they impact your results.
Thanks, that's Interesting, I will check that out, I have been tweaking a workflow I have made that I have been meaning to post, it can pay to tweak the saturation level of the images as it often comes out a bit high with this model, it also used Perturbed Attention Guidance and Lying Sigma's Sampler to really boost the detail, I will post it later.