Best settings:
Sampling Method: Euler A or Euler A SGMUniform - you can mess around with this
Sampling Steps: 20 ~ 60 depending on your sampling method
Hires. fix: Enabled
Hires Steps: 15 ~ 30
Upscaler: 4x_NMKD-YandereNeoXL_200k - You can test others.
Denoise Strength: 0.4
Upscale by: Preferably x2
CFG Scale: 6 ~ 9.5
Clip Skip: 2
Important:
Positive prompt must contain: score_9,score_8_up,score_7_up,
You can add "lineart, [artist tag], outline" for a more stylized result.
Negative prompt can be empty or contain the default pony negatives.
The promting format from Pony Diffusion V6 works well:
score_9, score_8_up, score_7_up, just describe what you want, tag1, tag2If you haven't already, I highly reccomend reading through the Pony Diffusion V6 XL model description.
I use Forge as my main UI because it has so many optimizations for running XL/Pony models on lower end GPU's despite me having an RTX 4090, generations are alot smoother. I'm not sure how this model performs on other UI's.
I haven't done much testing. If there is anything amiss, please let me know. Feedback is always appreciated :).
Post results and, most importantly, have fun!
Description
Better character context.
FAQ
Comments (18)
LOVE the artstyle!! Thanks for sharing the model 😍
Thanks for trying it out <3
@Mikoshii Are you planning to release an illustrious based merge?
I haven't really given it much thought, Illustrious is still new and hasent been refined to its full potential just yet although there are some pretty decent models released already, maybe in the future when there are more refined models on the market.
@Mikoshii Looking forward then !!
I personally prefer this model over EbaraMFCG. Thanks for sharing!
Hi Mikoshi, I'm having a problem with v1.1 of the model. I generate it, but the images come out with dull colors and no saturation. In the preview when the image is being made, it has the right colors, but the image is being delivered without saturation. I'm using your recommended setting which is in the description. I tried it without and with hires fix, but it didn't work. Below I'll send you some images of the settings and a photo of the problem so you can better understand what I mean.
Sorry for the late reply, try using sdxl_vae.safetensors or sdxl_vae_fp16.safetensors in the VAE dropdown.
@Mikoshii now this, https://imgur.com/fVgrPCH, I imagine it's a problem with my webUI, because I tested it on an SD in the colab and it didn't give me any problems (I use it on the Kaggle site)
@Lordland It's possible, maybe you have an extension active thats messing with the generation process? and what happens if you set the CFG scale to something like 5? It could even have something to do with your memory cache, you can try running RAMMap to resolve that https://learn.microsoft.com/en-us/sysinternals/downloads/rammap. https://civitai.com/articles/8060/how-to-clean-ram-free-up-ram-windows-official-microsoft-tool
@Mikoshii same result CFG scale 5, with the sdxl_vae the saturation problem solved but when i put a lora the model breaks again(last imgur link)
@Lordland Is it possible for you to try it on a different UI like Forge for example?
@Mikoshii I'm going to try forge, the colab UI was forge
@Mikoshii apparently now working normally in forge, https://imgur.com/zR55XT2 https://imgur.com/myV4yU7 , any new problems i'll let you know
@Lordland Glad I could help
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.