Reviews
Aurora
Overwhelmingly Positive
100% of 2,213 reviews recommend it · 2,213 reviews
mageofthesands
Can be a bit finicky on prompting, but results are good.
starrysheepi
Amazing model for landscapes! Very easy to use and has amazing results even for a complete beginner at AI art! :)
Hi, I've done a quick test.
Luna's models are always flexible, this is why i love you <3
It can do anime, realistic, and 3d without much hassle,
And every generated image looks good.
Iora also fits quite easily.
And just like kenshi, this model reward people who like to tinkering and trying different prompts.
Oh but i find that "tattoo" are hard to appear, eventhough im use lora.
You might want to add more image data of tattoo.
Yumeshiro
A fantastically versatile model, with the ability to generate images of all kinds in high quality. I love it. Luna did a great job with the description, so there's not much I can add in to help others with its capabilities except to say: Experiment and try new words and combinations! The first image in my review was made almost entirely with non-booru tags, and all of my included examples use no embeds aside from Aurora-specific ones (auroranegative, khfb) and/or sometimes easynegative/badhandv4.
There's a ton this model can do, and in the past week it's become my de-facto go-to anime model. It works fantastically with LoRAs and other networks, and can create tons of different styles.
RalFinger
Great fun to use, I also used the prompt from JTZ and modified it and added some of my LoRA´s, enjoyed it a lot!
Think I finally found the best anime model for SD I was always struggling to look for since months ago I first discovered the SD community as a whole when it comes to prompt accuracy and consistency. It's almost always accurate and consistent with any kind of prompt I come up with, more so than other models I used before such as Anything and AOM, so much so that I deleted both models the moment I tried it out and realized how truly versatile it is like the description here mentions.
It also works nicely with any kind of LoRAs as well, while as for embeddings, besides those that come with the model, I didn't have to use any negative one at all compared to AOM and Anything, I do still use a negative LoRA and hands hypernetwork in general tho.
My only two gripes with it are that I mostly get grainy backgrounds for some reason, putting 'film grain' kind of helps a bit but not much (does anyone else have this kind of issue when using this model?) and that it kind of struggles at managing colors most of the times, only sometimes nailing most or all colors in a prompt. It also rarely screws hands too in certain generations, but not always thankfully.