Today I took a mission to generate cute cat gangster AI art images. I started using ImgCreator.ai. It’s an AI art generator based on Stable Diffusion but modified for better realistic photos and 3D rendered objects.
The theme is cats that look like gangsters. I mainly used ImgCreator.ai which delivers excellent results and then also used 3-4 images from DALL-E which did better with the clothing and holding a cigar in hand.
I wanted hyper-realistic results. I wanted the cats to appear realistic. In ImgCreator.ai, I went with the ‘realistic photo’ category, ‘backlight’ for the light, and ‘wide angle’ for the camera lens. All those settings have further optimizations so it seems. This yields excellent results in most of the AI-generated images.
Prompt I use (with a few changes between images):
“gangster criminal Cat wearing expensive gentlemen suit clothing, bow tie,gangster fedora hat, wicked evil face, smoking a big lit-up Cuban cigar,smoke coming out of the cigar, evil eyes, nightclub scene, shadey smokey scene,hyper realistic”
Here are more images generated mostly with ImgCreator.ao (modified Stable Diffusion) and DALL-E.
Overall, I definitely preferred the results from ImgCreator in that particular art direction.
I wanted wide-angle portraits, dramatic lighting, shady nightclub environment lighting, cigar smoke, and gangster clothing and accessories.
DALL-E was good with the clothing, ImgCreator.ai wasn’t. But the cigar location in the mouth was perfect on ImgCreator.ai, whether with DALL-E I needed to be more accurate with the prompt description. Even then I didn’t get great results as I got with ImgCreator.ai (modified Stable Diffusion).
ImgCreator.ai devs mentioned it’s optimized better for realistic photos. Playing around with the cat gangster idea with many prompts, it definitely seems that this claim is true. Or better to say, it works to my satisfaction better than DALL-E and way better than Midjourney –test (photorealistic parameter). Midjourney was actually the worst for that based on my experience.
To my knowledge, Midjourney uses Stable Diffusion for the –test parameter to generate more photorealistic images, but it didn’t give me the results I was looking for, far from it.
So first impressions with ImgCreator.ai for certain AI art artworks are positive. I’ll play with it more and share more things when I get more experience with it.
More about ImgCreator.ai in this video.