GPT Image 2 from OpenAI on low quality setting. Excellent value with strong prompt fidelity and text rendering. Supports prompt-to-edit and multi-image fusion.
Model specifications
Art
Realism
Cost
$$$$$
Prompting
Typography
Recent creations by the community
See what people have made with this model
0
0
The Chimney Sweep's Stand
The Descent of Ash and Iron
While the model has good prompt following, it leaves a repeating light/dark pattern, like ripples or dots, over nearly all images it creates. At times it exacerbates this to an excessive degree, with only one character in focus, and the rest of the image an unrecognizable rippled blur.
I get a lot of blocky pixelated noisy backgrounds. The higher rendering resolutions look worse. Requires extra prompts to look good that the other algorithms don't. Even the old algorithms can look vastly better under simple prompts.
Plus the auto generated Vincent prompts normally fail because it references people's names.
Currently would rate this low, and I wish I could hide this model
Detailed and interesting output but looks like prompts should not have any names (even for style) or it'll be rejected... for people, I feel like you also need to describe clothes well so it'll not be rejected
Comments