Setup · 2:12
Fluxmania setup
Photorealistic images with a simplified workflow.
- Fluxmania targets photorealism.
- Simplified workflow: prompt + dimensions.
- Load the model.
| Input | Use |
|---|---|
| Prompt | Description |
| Dimensions | Size |
Do this in your own tools, step by step. Ticks are saved on this device.
Model?
Fluxmania.
Inputs?
Prompt and dimensions.
0:00 welcome to episode 41 of our comfy UI tutorial Series today I will show you how I generate photorealistic images that look like real photos using comfy UI I have been searching for a model that lets me create realistic images and today I found one that is getting close to what I want it's called flux Mania version 5 and it was made available for download today when I recorded this video this model is created by merging different checkpoints based on the flux Dev model So in theory it should be a little better than the original flux Dev it has two models one that includes clip vae and all that which is usually used by people who use Forge UI and the other is the smaller fp8 version which doesn't include clip and vae I'll use this smaller version and add the models that work best for my system open comfy UI and from the workflows menu you can open one of these workflows I will add them to Discord for free as always it's a simple workflow and I've included everything you need to get started in this note to begin click here to download the model in the models folder go to the diffusion models folder and
1:09 place the model there as shown in the instructions do the same for the T5 model and place it in the clip folder I'm using an alternative clip L model here which is supposed to be a little better than clip L of course you should already have the flux vae if you've used flux before there's only one custom node to avoid errors so go to the manager then custom node manager search for easy use and install that node after that restart comfy UI once you've added the models and the node you should be able to see the flux Mania model in the load diffusion model section and the clip models in the Dual clip loader now you can add your prompt set the image size and run the workflow what makes it realistic is the prompt and the model which seems to be trained on more realistic images I also added a clear V node here because sometimes after running flux a few times it might get stuck this node helps it run more smoothly and this is the result pretty realistic isn't it if the workflow works well for you you can also try a compact version of this workflow I added the word compact in the workflow name it's the same workflow I just merge the nodes
Get the other 3 chapters, the exam and your certificate for LinkedIn, plus all 700+ AI courses.