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
FAQ
Comments (17)
Probably, but I don't speak their language. I always thought that the Asian school systems were already very good and strict, so most Asians should already have a great education. What more can they be trained on?
@6tZ ⊙ ⊙ Asian face, there are many beautiful women in East Asia. Japan has many beautiful AV actresses.
@cai567890 Yeah but you need to speak with these women to train them. I cannot be held responsibility for educating and training the Asian people.
I am missing JDC_ResizeFactor and the Face Detailer Appended Prompt and do not get it work. Any idea how to solve that?
Doesn't the manager tell you what you need to get?
You can just use any concat/append/string combination to get the same result.
Face detailer, search youtube for some tutorials, there's many versions of this as well.
V6 and 5.2 have an awful prompt adherence, Danbooru tags knowledge and LoRA compatibility.
Forcing natural language prompts on any ILXL checkpoint is a no go for me. I want to prompt using Danbooru tags because i m aiming for accuracy, not luck. Using natural language on ILXL is like flipping a coin and hoping it lands on the side you bet on: it's neither reliable nor accurate.
I don't know about older checkpoints, but testing the two most recent ones and getting these results has completely dispelled my desire to try others.
Fair and I agree with the feedback. The Spectacular series is more of what you want, but is much less realistic in nature.
Additionally, you can use this model as a second pass (img2img or later steps), to get a realistic render on your previous prompt-adhering start. Some great things can be done like this.
well i tried 6.0 and was very suprised how good it was, so maybe depends on the task
Amazing. These get worse with every new version
Which is the best, anyone tested all modeld 1-6.0 ?
I have! I posted some images for each model so you can see them and judge for yourself. I recommend checking them out. You can find them in the card for each model.
@6tZ thank you for your reply, i watched your gallery ofcourse. But i ment 1:1 Comparisons (pairs of imanges or goups of exact same prompts like XYZ Chart some other Creators sometimes post) i asked since i am on limited Bandwidth. I already extracted intel from the comments that Version 4.0 is good. And i tried 6.0 which is very good from what i tested so far(20 Step DPM++ 2M automatic 1280x1024), it suprisingly works without hires reasonably (many other modes look very noisy without highres, i thought this noise was normal until now)
@pink0909 That's fair.
I haven't done a test like that, I mostly look forward, but if anyone makes it, send me a ping :)
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.