This is my first test as a NoobaiXL mix.
I mixed some of my favorite models with the intention of a model being capable to do a varied kind of styles and body types, mantaining the capability of character recognition of the last epsilon version of NoobAi.
The settings I use:
Vae: Not needed
Size: 1334 x 768
Sampler: Euler Dy
Scheduler: Beta
Steps: 25-30
Cfg: 3.5-7
Upscaler: x1,5 size, 4x_foolhardy_Remacri 12 steps, 0.45-0.5 denoise.
Prompt and Neg Prompt: (From here:https://civarchive.com/articles/9695/f-best-negative-prompt-for-noobai-xl) I followed the example from there and I liked the results.
I reccomend also: https://www.downloadmost.com/NoobAI-XL/e621-artist/ to look for characters and styles.
Description
1.0, first testing version.
FAQ
Comments (9)
I'm glad to see a gay, male-oriented model! I probably won't be trying it due to commercial restrictions usually leading to problems later, but I hope it will lead to more good men models.
I don't really mind if anyone do it. But I thought I had to put it since the original noobaixl has it in its license too and this is a merge of different models based on it.
As an avid fan of HMenMixsd1.5: this is a gift of divine origin.
Thank you FMen: now I have finally been convinced to learn how to do stuff w NAiXL, left behind in my toolkit so far cuz it crashes my a1111 all the time.
HmenMixXLNAIXL works well on a1111 so far using Euler Beta. Even after updating I haven't found Euler By.
I'm excited to poke this new model into submission
🤜🤛
I use reforge on stability matrix, a program that automates the instalation of a lot of AI image generation programs because is really easy to use and I'm lazy. Reforge is like a1111 but should be a bit faster and it's constantly updated with samples and schleduders to experiment with. The only downside is that it's not compatible with a lot of plugins. But if you don't know it I reccomend to give it a try.
This is looking like my new go to model… thanks! ✨
This is now my goto favorite model!
Not only it always generates the characters well without making it distorted, this Checkpoint seems always "Blends" well with all lora I use! Unlike other checkpoint that keeps making lora character too muscular and such, this model seems always generates it just like the dataset used!
Thank you for making it! Looking forward if you gonna updates it!
Tysm! The thing is... I totally want to since I discovered a lot of models I love to merge to improve this since I made this one. But just the same day I made this my hard drive just explode (I had to buy a new SSD), and I suspect merging had the fault of it, maybe I messed up somehow, and now I'm a bit scared jahsjd. But I totally want to, I'll probably inform myself better before doing it or try to do it online if it's possible.
Nah, dont worry~! Sorry if what I write seems like pressure you to update this checkpoint. This checkpoint is already great as it is, and I just want to appreciate what you done. Just take it easy~!
@RaidenFan86 nah really, thanks for encouraging me and I'm glad you like this