Get the pro version of my checkpoints and workflows on my Patreon.
Onsite generations are permanently available on these models:
👉 Realism_Illustrious_By_Stable_Yogi : https://civarchive.com/models/974693?modelVersionId=2091367
Read "About this version" Section for Model details and settings.
Important Usage Tips to get similar results like my images.
Add Realism_Illustrious_Negative_Embedding at the beginning of your negative prompt.
Add Realism_Illustrious_Positive_Embedding at the beginning of your prompt.
Join me on my Patreon for exclusive perks and early access to unique resources.
To discuss custom Lora's or models, feel free to connect on Discord.
👍 Like this model to keep me motivated and inspired to create more!
💬 Drop a comment and let me know what you'd love to see next.
🌟 Review this model to help me improve and make even better creations.
🔔 Hit that notification bell to stay updated with my latest models and updates!
Description
This version combines V1.3, V1.6 and V1.8 and other minor tweaks to enhance realism and lighting, prompt adherence is better and on par with V1.3.
let me know your feedback.
FAQ
Comments (5)
V2.0 definitely best version so far, especially for generatding more than 2 characters (but needs afterdetailer),
Thanks for the feedback.
v2 is a great model and my current favorite model on all of Civitai.
I think the face distortion problems are a fundamental limitation of SDXL and its VAE, so face A-detailer will occasionally be needed.
I experimented with various accelerators, including Lightning, Hyper, and DMD2, combined with samplers such as LCM, Euler A, and DPM++ 2M, using the scheduler SGM Uniform.
After testing, I found the best setup to be the DMD2 accelerator paired with Euler A, a CFG of 1, and 8 steps.
Mixing the DMD2 LoRA into the model increases generation speed.
Using a CFG of 1 significantly increases generation speed—approximately doubling it compared to higher CFG values—at least on DrawThings. I’m not sure if this speed boost applies to other apps like Comfy or Forge.
With the time saved, you can comfortably generate images at a Hi-Res setting of 1280x1856, using a denoise value of 0.4.
Summary: This is a amazing good model.
Thanks for the detailed feedback.
DMD2 works best with Euler and CFG 1
And that will work on all Uis.
I will be uploading the DMD2 version soon. Its under testing.
All those accelerator LORAs are only really good if you want to reduce compute time by reducing steps at the expense of quality, especially if you have a weaker GPU. The other thing is that base SDXL pixel size was trained up can be pushed past to a point, but it's still going to have an upper limit before upscaling techniques are needed to push past, with resultant quirks and quality trade-offs.