The new version of the image generation is so incredibly garbage it's genuinely unusable. Instances of AI artifacting have multiplied exponentially, the system barely follows prompts, characters are jumbled messes of blurry features, it's awful. If it's an issue of expense, I would be willing to donate along with countless others if you took a collection. Please as a community we beg of you to change it back. I would be happy to pay for access to the better model, or a subscription for improved functionality, just anything but this.
I'm using the exact same generator that I've always used. I work with AI every day, and I've changed my prompting strategies multiple times throughout this change to try to give the engine the benefit of the doubt.
"Digital illustration with a painterly, tumblr webcomic look. It blends anime-influenced character design with Western indie-cartoon expressiveness. Shading is a hybrid of cel and soft airbrush. Planes are defined by clear shadow shapes, then feathered with soft brushes. Unique photo-quality visuals in high definition.
A man with pale skin, black hair, and fangs dressed in a stylish late 2000s outfit consisting of an open leather jacket over a white tank top, silver facial piercings, and black skinny jeans. He is wielding a glaive that emits a smoky, silver aura from its blade. He is standing on a stage, surrounded by a crowd of thousands.
Closeup of the character with visible facial features, composition matches that of late 2010s era tumblr art posts. Serious, non-cartoonish features."
Is my exact prompt. It's on par with many other prompts I had been using before, now adapted to the new scheme which relies less on negative prompting and () emphasis indication. I pretty distinctly ask for something specific, and define the parameters for that exactly. I state directly that the image should not have exaggerated cartoonish features, and yet I get this fucking slop.
I can't use any of these images. No combination of prompting principles seems to fix this issue. If you genuinely think that is better than this series I was able to make with the previous generator:
You are genuinely mentally deficient. Your "help" would change nothing.
Edit: One of the site admins also directly posted that the image generator is empirically downgraded for the time being because the text model is taking up financial resources while they debug it. My, and many others' offer is genuine. If the site needs more money, I'd be happy to provide some for the same quality I'd been getting before as quite frankly, there was nothing else good out there for non-scam prices. This site has a genuine interest in generative media and I'd love to support it. Even if it's just a donation link. So, kindly shut your fucking mouth.
Double Edit: Inb4 you say some shit about "OH WELL YOU HAVE THE WORD CARTOON IN YOUR PROMPT" western-cartoon indie is an art style. One which doesn't necessarily have the proportions of cartoon characters, but rather defined by solid colors, hard lines, and complex character design with a few defining traits for each character. Think like Avatar: the Last Airbender. That would be an example of the style. But just to make sure you understand how little that matters, I did another generation omitting that entirely in favor of defining the facets of the style one-by-one. Here was the results:
Still cartoony, warped, barely recognizable slop.