The module already has VAE. Unless you have specific coloring requirements, it is recommended to use the VAE inside the module. Currently, personally recommend NoobAI ( 1.7, 1.9, 2.4, 2.8, 3.3) . If the FP32 upload fails multiple times, stop attempting to upload.Buy a Coffee for me☕️
Suggested: Txt2img 720p+ with Hires fix. Upscaler recommended Remacri setting to 1080p+.
Prompt Prefix:
masterpiece,best quality,perfect quality,absurdres,newest,very aesthetic,vibrant colorsNegative Prompt:
shiny skin, shiny clothing, worst quality, low quality, lowres, signature, username, logo, bad hands, mutated hands
V3.0, V2.9 Variant suggested: forge and use latent modifier integrated
Suggested Resources: Eyeful, adetailer, remacri, Face_finder
Simplified instructions for the standard guidelines of NoobAI
Character and artist tags Prompts achieve reliable performance when applied to NOOBAI CSV files with the Danbooru+e621 dataset exceeding 300 images in the official records. ( NoobAI huggingface csv).
The illustrious checkpoint pre-release beta version for testing must be used in accordance with the illustrious XL standard usage guidelines.This illustrious checkpoint is not stable, and the background sometimes disappears
Description
Recommended:
Sampling method: Euler a
Sampling steps: 13~25
FAQ
Comments (14)
Dude, you are on a freakin' roll with this. You are fantastic!
不知道为什么眼睛总是有问题，用了推荐的detailer还是崩坏，可能是我本来就不会用detailer（？
你可下载forge后贴上我的示范图直接用我的示范图，我的示范图都有保留生成内容，用图片资讯将图片资料直接带到文生图后打开ADetailer
@ok_furry_image 抱歉我有点没明白 我是本地的 一样的吗
@iojw69463 是的,我都是用forge本地生成的，可以直接用圖片資訊做文生圖 https://github.com/lllyasviel/stable-diffusion-webui-forge
@ok_furry_image okok我用的是秋叶 我下个forge看看
@iojw69463 The eyes are a common flaw with this type of model. You’ll need to use a dedicated eye LoRA to improve them—or avoid generating full-body images
@Darkness007 got it,thanks
@ok_furry_image 您使用过regional prompter这个插件么
@ok_furry_image 我的forge已换了适配它的reagional 但是它总是不起效果 尤其是我想分割Lora的时候 这种情况怎么解决
Testing your V3.2 Model in the 13gb version tonight.
In summary: The best illustrious model I've tested so far.
This model produces really high quality artwork, with an amazing sense of composition. It is superior to anything I've tested in the illustrious family.
It understood very (VERY) long natural language prompts for my artwork really well, and it has a sense of composition that is shocking, honestly.
It only was matched by really well prompted anima shots, but anima being a slower model, and harder to control, had to have many more generations to get those really great ones that this model just did on every other 2, for half the time Anima takes.
As with any illustrious model, it seems to be very versatile too, and does an OK job at upscaling, although i still find a couple other models to have better refining qualities in my test, this exceeded everything I've tested again in pure generation.
This model is a great place to stop and do your entire generation flow on it. No more loading two completely different weights. Just load one, and do it all. I'll continue testing in my actual work, but it seems like a great place to stop for a while until Anima comes.
Furry content is not really my cup of tea, but I know where to look for great models and it's always you guys :)
I'll be sending a tip for the work done here. Congratulations.
Would you mind if I asked you how you test a checkpoint?
@sirochan4646 If you're asking how i came to the conclusion, upon comparison of same prompts & settings, this model came out on top for my liking in 72.8% of the cases, after testing the 13gb version against my preferred models.
That is the exact number after comparing manually around 994 images, 142 images per model vs previous favorites models, and picking a winner per prompt.
Think of it as, For every 10 images, 7 of them i picked this model, out of 1000, 700. More or less. It's a very strong sign imo.
It adapted well with loras, different styles, and the general composition always worked good, especially in cases with a lot of natural language. It's especially interesting how strong style lora effects have on this model, which is usually the opposite with many checkpoints. It does especially well in single & dual character compositions.
There are no Magical solutions for SDXL as you might know, but this one is quite good at what it does, given SDXL's limitations.
In the open source space there's the new Anima models which do very well, even if they're a bit more complex to work with and don't have a strong sense of styling as of yet.
@BoundingBoxes Run the same Settings with different checkpoints... Thank you for sharing :D