I decided to mark three clear lines in my realistic models, the SatPony model focused on the purest realism, the SatPonyJourney model focused on having the maximum amount of details and this one, the SatPonyNoScore model that focused on lighting and textures, you can use score markers or not to your liking.
The VAE is not baked, use SDXL_vae
Clip skip: 1
Upscaler: Im using 4x_NMKD-Siax_200k, upscale 1.5 Steps 10 denois 0.35
Width and height: 896*1152
CFG Scale: 7
Steps: 35
The sampling method i use is DPM++2M SDE Karras but try with others if you want
06/30/2024
Some people mentioned to me that they had problems with the VAE so I decided to make a new version with some improvements and that it is BakedVAE
This model started trying to be a version two of my previous one but evolved enormously thanks to the fact that it no longer needs Score to create beautiful and realistic images, like my previous model this one specializes in realism, several things are different such as more realistic, better colors, better textures and better understanding of prompts especially NSFW, Feel free to always try it with respect for the community
06/27/2024
I am doing an experiment with img2img, it was not planned for this to happen with the model but after some tests to convert anime to realistic photo and with the permission to use the images of two excellent artists like JulzzR and fortylove I came to one amazing results, I hope you can also use this tool
Description
The examples I made were made with the DPM++2M SDE Karras sampling method (you can use DPM++3M SDE Exponential), with 35 steps, the size 896x1152, the Hires.fix with the 4x_NMKD-Siax_200k upscaler with 10 steps, scaled to 1.5 and denoising 0.35, feel free to try other configurations
FAQ
Comments (25)
Is the VAE baked in?
I use an external VAE
Ahh OK, I didn't realize I had sd_xl_base_1.0_0.9vae.safetensors selected for regular SatPony - using that on SatPony No score. Great models though! Keep up the good work!
Wonderful masterpiece!
Ty for your review, it's worth the effort for this words
why you have to push your model with 5 loras OMG
my model is better with-out any lora ^^
The examples in the main publish are without Loras and are very good, I like to try different Loras because in that way many people can view the images
By the way, I think your model is great and I'm not try to compare with anyone, in their own way every model do a great job
thx , all fine ... every one try its own way
This model is insane. Out of the box amazing pictures.
Ty four your kind words
@Sateluco Of course! I have been experimenting the last few days and I did notice that there are times when the bodies and skin will suddenly turn "plastic" looking. I'll have to pay more attention to see what triggers are doing this, but it will suddenly lose the photography style. It tends to be with longer prompts. The shorter, more simple prompts give much better results it seems.
@fantasyai449208 exactly, maybe i have to mention a couple of things that work better for me and that is shorter prompts and NO negative propmpts or very puntual, thanks for your feedback
@Sateluco hello there! I have been using this model for about a month now. Absolutely love it. I am not sure what you have planned next, but a mix of this with Fennfoto or Vendo Pony would probably be amazing. This model out does them in prompt adherence, but only gets about 80% of the skin textures that those do. Just some additional feedback. Whatever you do, keep up the good work! I love the new Journey model btw. Thanks so much
@fantasyai449208 Ty four your feedback and your words, worth the hard work to share new models
Great model. I might even prefer this over ponyRealism, but it's a close contest. Will experiment more.
Ty for your comment, I hope to see your creations
I am doing an experiment with img2img, it was not planned for this to happen with the model but after some tests to convert anime to realistic photo and with the permission to use the images of two excellent artists like JulzzR and fortylove I came to one amazing results, I hope you can also use this tool
where's the sample? ,
@nicetry20010 yo can check it in the examples of the previous version, they are pinup
The VAE in the model is not available
I'm testing your checkpoint. What's your config for ADetailer (faces)?
i barely use ADetailer but when i use it is face_yolov8n.pt with sampler DPM++2M SDE Karras