I played with different NN (ai image generating) models, and I must say that so far, I'm HUGELY impressed to the point that I would say that these things will drastically affect how artists will work in the future.
Breaking down all available options, you have 4 choices:
• DALL-E 2
• Stable Diffusion
• Different apps
DALL-E 2 is interesting and good at generating 'impossible' photorealistic things, but it now feels a bit outdated.
MIDJOURNEY is the most promising model right now. It is simply much better than anything else out there. Photorealistic or not. Although its previous render model was way too obvious MJ-like recognizable render.
STABLE DIFFUSION is something in between of MidJourney and Dall-E. I wasn't impressed by it so far. It feels like a slightly upgraded DallE. However, people manage to make some fantastic stuff with it.
APPS are primarily utilizing limited versions of the previous version of the Stable Diffusion build and are... Honestly, crappy.
Here's my basic prompt (with slight variations)
“gate to the hell, cinematic, realistic, hyper detailed, ultra HD, unreal engine 5, octane render, photo realistic, --ar 9:16”