Now, this app is really cool. First, you can make (almost) exactly what your thinking. Also, you can know what your child would look like with this app! Now, there is some downs. There is A LOT of ads! And sometimes theres so much ads that the app cant handle it and it freezes, forcing you to go out of the app and close it. Also, please note that there is a lot of things you need premium. For example: avatars. Now it might not sound like a lot but let me tell you, theese avatars are take SO long to find the 10 pictures of ONE person, and then you gotta select 10 DIFFERENT FILTERS! WHEN THERES ONLY LIKE 13! And when you find out its premium, you get so mad cause you just spent so long just for them to tell you to pay for your work! Its kinda crazy that they make you do that. If you want the avatrs, then im sorry to tell you, but your gonna need a better, different app. AI Arta, if your reading this, please improve on your app. Its getting kinda annoying. Anyways for all of you other people thatnread the whole thing, congrats! Cause now you know what to expect.(unlike i did)

I had purchased the year subscription for this app. For a while it was decent enough to get close to the image I was looking for. Some time later, they updated to SDXL 1.0. I was getting better images, with the weird extra limbs here and there, but overall, I was happy with the images. Then I started using the photographic filter. The images were good for a time, I was getting what I wanted, but now, yikes. The photographic filter especially went down the drain. The images look blotchy, like they were painted on in dabs of paint and the images are blurry. Don’t even get me started on the other models, anime? Laughable. Cinematic? Even worse. Luckily, the regular SDXL model does a bit better, but it blurs the backgrounds and the models are not as good. All in all, while the SDXL model is still working for abstract images and skylines, the photographic filter (and some others) is terrible. I have cancelled my subscription.

When trying to use an image as reference for a new image, this algorithm fails. No combination of prompts and reference strength generates an image even close to the desired output. If you’re looking for “take these people and generate an image of them that is similar” this app can’t do that. More bothersome is that this app can not render a curly haired person without making them black, and it can’t make a black person with straight hair. It doesn’t understand that some white people have curly hair and some black people have straight hair. Edit: the developer response is to change steps, reference strength, and other fields, and to toy around with the character limit. My point was that even changing all of those things many times, no output was worthwhile. For example I was asking for a family portrait style photo, with a man, woman, and two daughters. I described the individuals. I tried various levels of specificity in describing them as well. It kept rendering a gay male couple with two Victoria’s Secret model-looking daughters. It was funny, but not useful. And there wasn’t any way to provide feedback to assist the AI in learning. Again, entertaining but that’s where this app ended for me.