Another new model, this time focused on realism but detailed, most of my models are focused on reaching the maximum level of realism, I added the term magic because the amount of details that can be achieved are magical, something that I was experimenting with very good results and that you can try was to add the clip_l in the VAE, it totally changes the equation and sometimes it can give extremely surprising results, as always I hope you share your images since it is the only way to improve, thank you.
The VAE is not baked, use SDXL_vae
Upscaler: Im using Remacri_original upscale 1.5 Steps 10 denois 0.35 Hires CFG 5,
Width and height: 896*1152
CFG Scale: 5
Steps: 35
The sampling method i use is DPM++2M-SGM Uniform
Use this model, taking care that it has a great chance of creating NSFW images.
Description
In this new version I think it has improved a lot in every way, much more realistic, better adherence to the prompt, greater variety of ethnicities, try some prompts with CLIP-L, some really have a substantial improvement although not with all, as always I hope you share your creations, as you know I do not like to put my models in early access but any tip is welcome, thanks for uploading your images, it is the only way to improve.
The VAE is not baked, use SDXL_vae
Upscaler: Im using Remacri_original upscale 1.5 Steps 15 denois 0.35 Hires CFG 4,
Width and height: 896*1152
CFG Scale: 5
Steps: 35
The sampling method i use is Euler DPM++2M-SGM Uniform, but sometimes i use DPM++3M SDE-Simple
FAQ
Comments (11)
Great model so far!! Tested with Accelerater+ILL (LLORA), 16steps CFG3, Euler A Exponential+SDXL_VAE with good results!
It seems I can now delete all my other realistic Illustrious models from my computer.
Amazing update.
why does every gen have butt chin?
Thank you very much for your support, it always gives me the cheer to create and share my work. Look forward to a new checkpoint soon, which I think is one of the best I've done.
@Sateluco People like you who make good checkpoints that we can all enjoy deserve all the support. So far I have only used the 2 mentioned, but I also want to try some older ones like the 'Illustrious-SMRealism' and the 'SatPony-JourneyV3'. When you are working on a new checkpoint, that is what I call good news to look forward to! ;-)
Noticed this long time ago. When using image to image, this model always tends to change the face to the one like on first example picture
I'd like to see a v3 of this model. It's very good and is underrated in my opinion.



















