This workflow allows you to make photos created with IntoRealism Ultra, as well as other SDXL and DMD2 models, more realistic.
It also includes an upscaler and a facedetailer to enhance realism. These options can be disabled if desired.
It is possible to make it even more realistic by increasing the Denoise, but the main image will be slightly altered.
You can download z-image here : https://civarchive.com/models/2168935/z-image?modelVersionId=2442439
Description
FAQ
Comments (13)
in comfyui which costum_nodes verson did you use? Cuz the workflow in the middle freezes my comfyui server
I'm a total noob, just learning comfyui, added your workflow, everything seems to work, but where do i plug in my own image that i generated? sorry if dumb question.
You want img2img from a photo that has already been taken. That's different. But I'll make a WF for that ; )
I'm trying to get more detail out of the final image. Do you think it's a bad idea to use an upscale model like ERSGANx4 instead of a normal upscale without a model?
All the workflows I see do an upscale without a model. I've always done it with upscale. I don't understand why Z Image Turbo doesn't respond as well as other models with upscale.
I think it does respond well with upscale. I've generated some images with 1080x1920 as start resolution then do a 4x Ultrasharp, then a Lanczos to 1632x2880 and do a 2nd pass on that with 4 steps at a denoise of 0.3 . Still experimenting with this but it looks fine to me. See e.g. here, added a comparison for the small goat in the back of the image: Image post by gman_umscht | Civitai
@gman_umscht Yes, I see that the finer details have come out well.
The problem I encounter is because I have never used distilled models, I have always hated them for the few steps and the CFG at 1 without negative.
From the tests I've done and as I can read from your info, I believe that distillates give their best in a few steps.
Thanks for the example.
Where am I doing a mistake:
Every prompt I enter leads to the same image?
You need to replace fixed with randomize in your WF.
@enzino thank you for your reply. I must be blind, still not finding it. As to make it clear: I always get a woman in Bikini, even if I type "dog" in the prompt field. The woman changes, but never getting something I prompted.
@SaBaDaBaDo Empty Base Prompt Clip Tex
@enzino Thank you a lot!
Маэстро,чекпоинт ZiT готовишь? Или базовую модель подождать?
SO many problems.
ComfyUI-FRED-Nodes_v2 - 404 error!
RVtools_v2 - This repository was archived by the owner on Oct 8, 2025. It is now read-only.
LocalMediaManagerNode
Any to String (Image Saver)
Image Multi-Switch [RvTools]
in subgraph 'Generation Preview'
Latent Multi-Switch [RvTools]
in subgraph 'Generation Preview'
Model Switch [RvTools]
in subgraph 'Generation Preview'
Installation Error:
Failed to clone repo: https://github.com/alexopus/ComfyUI-Image-Saver
I have not been able to run this so far due to those errors and issues.
Also you have a node Local Media Manager. That does not exist. I cannot find it.