version 3.0 aims to have both "photorealistic" and 2d capabilities,
my photo realistic settings are
sampler restart 25 cfg 3-7
score_9, score_8_up, score_7_up, realistic, irl, photo, realism, ASCII Masterpiece, photorealistic
score_6_up, score_5_up, score_4_up, text, shiny skin, watermark, low-quality, signature, moiré pattern, downsampling, aliasing, distorted, blurry, glossy, blur, jpeg artifacts, compression artifacts, poorly drawn, low-resolution,
hiresfix 15 steps denoising strength of 0.4 1.5x with nmkdSiaxCX_200k
my 2d settings are
Euler A 25 cfg 3-7
score_9, score_8_up, score_7_up, source_anime, 2d,
score_6_up, score_5_up, score_4_up, realistic, irl, photo, realism, ASCII Masterpiece, photorealistic, watermark, text,
hiresfix 15 steps denoising strength of 0.4 1.5x with nmkdSiaxCX_200k
i use adetailer to then refine face, i highly recommend using an eye lora at this stage to improve the eyes more.
Description
FAQ
Comments (5)
Thanks for the work, will be playing around with it during the weekend.
I'm so glad you like it! I would love to see some of your gens posted here!
If the guys who captioned Pony's dataset were to work for Stability AI, we'd finally have a base model that actually works. PonyXL v6 is really the most amazingly well trained generative AI ever made.
Thanks for working on this interesting alternative.
When you say to use that zPDXL tag as a base prompt, do you mean to say that you baked in zPDXL, or that it assumes that we have that downloaded, and may need to rename it based on how we named the file? It's not listed in your example images as a resource, so I think it might be baked in?
It is not baked in and you have to download it i have linked the file in recommended resources. I will have to change the resources used.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.