Welcome to Perfection Realistic!
This is an Illustrious XL merge with a bunch of hand-selected checkpoints and LoRAs to improve the base model's realistic capabilites.
And wow, does it deliver!
Also available on mage.space.
It's focus is on creating high quality, photorealistic, but slightly exaggerated images with a focus on women.
Do not use the Normal scheduler. Use Simple or Beta!
Use euler_a, dpmpp_2m_sde or dpmpp_3m_sde as your sampler.
To use the faster low-step settings, you need to use the DMD2 LoRA. This allows you to use 8 or 16 steps and generate very fast. Remember to generate in 2 passes, either using "Hires fix" in Forge/A1111, or by creating and chaining a second sampler node in ComfyUI. A sample workflow if included in the sample images of the model. Use this as a reference.
Each sample/preview image contains the used workflow. Here's a quick article with simpler more beginner-friendly workflow. This is a recommended starting point.
Recommended Settings
Base Generation
Steps: 24
CFG-scale: 4
Sampler: DPM++ 3M SDE
Scheduler: Simple (beta upscale scheduler)
Upscale/Refine
Steps: 8
CFG-scale: 6
Sampler: DPM++ SDE (dpmpp_sde)
Scheduler: Beta
Denoise: 0.35
Description
I went back to version 1.0 and merged with a different WIP model as well as some other LoRAs. Greatly lowers the chance of young looking people.
FAQ
Comments (22)
I can't control her expression...whe she always smile? i wanna she cry -_.-
It can be a bit hard to control this model.
It does photorealism well, but fine-grained control = not as well.
Try increasing the weight of your expression.
(crying:2) for example.
does this checkpoint work with flux trained loras?
No. Flux LoRAs only works with Flux.
Does it work on the MacBook Pro M3 Max without CUDA, using Metal instead?
Not sure. If SDXL works there, it would work.
@6tZ yes, it's working
Does this work with Fooocus? I dont seem to be getting the same quality of image
I haven't tried it, but can't see why it wouldn't.
Regarding settings, you might need to use more complex setups, like a base pass, and then upscale pass to go to higher resolutions.
After a couple generations, I am blown away by this model. Great work.
Question, I am new to this. Can you tell me if its possible to train a lora on this model?
And if I download it, do I have to download another program to use it? What is different from downloading instead of just generating the images here?
Sorry if these are stupid questions.
You can train a LoRA on most models, this one included. It is generally recommended to train on the base model, listed under "details" above, although it depends on what you're trying to create. If you want to train on realistic images, you may have more luck training a lora on this model rather than the base Illustrious model.
-You'll have to download and set up another program in order to use the model outside of this site. Stable diffusion web ui and ComfyUI are probably the most well known, but there are others out there. If you decide to go this route you'll have to have a decent graphics card. That's also the main difference between generating here and locally on your machine. There are lots of tools, workflows, and ways to customize how you're generating images once you get over the hurdle of setting everything up. Generating images on websites is meant to be highly accessible, but the sites often limit options for the sake of creating consistently good images.
Good luck!
As SleepyDreamy said (thanks for answering)!
You can train LoRAs on most models. This one included.
After you download it, you need to use other tools to generate images with it, or train with it.
For generation, I would recommend ComfyUI or Forge.
For training, I would recommend Kohya_SS or OneTrainer.
Generating on CivitAI is easier since you don't need to install the tools and bother with "complex" setups. You can get started right away, and you don't need a GPU.
But it comes with a "per use" cost in buzz.
Using it offline, on your own GPU means you have greater control, but it might be slower, and consume electricity for your computer.
@SleepyDreamy Awesome. Thank you so much for the info. Greatly appreciated!
@6tZ Great info. Thank you for getting back to me!
What? is this FP8 model? Don't lie to me!! :)
Amazing GREAT!!
I dunno... you're forced to pick something when uploading the model. I have no idea!
This is so good! Even though it is focused on women (which generate beautifully btw), it still generates attractive and diverse looking men. Not to mention it generates good male body anatomy and good male genitals as well (and hands too!). Thanks for this awesome checkpoint!
I think it could use some male improvements. I'm not sure how diverse it is. I seem to get pretty generic looking men if I don't prompt for them heavily.
Thanks for the kind words!
@6tZ Hmm, I will upload some generations of males that I have done with this checkpoint. I look forward to further iterations/development and what that might entail. Most Welcome
@reetak Nice! Sounds good, please do share, looking forward to seeing!
Wow! This is your best model yet! Great work! Can't wait to try the 2.0 version on the on-site generator.
super precise