Model guide

Image generation model comparison

Two actual results per model using the same Chinese prompt, with parameters listed alongside.

On this page

Settings used

Text-to-image, one image per task, no fixed seed and no web search. Parameters are listed for each model; this is not an equal-setting overall ranking.

One prompt, two results

Two actual results per model using the same Chinese prompt, with parameters listed alongside.

Prompt

Generate only one square, high-quality commercial photograph of a quiet, bright modern cafe. Do not create a collage or multiple panels. Shoot at eye level from the indoor entrance. In the center foreground, place a light solid-wood round table holding a white ceramic coffee cup, a clear glass of water and an open magazine, with two simple green upholstered chairs beside it. In the middle ground on the right, place a tidy wooden coffee counter with a silver coffee machine, several clean cups and a small bunch of white flowers. On the left, show a large window with softly blurred but natural green trees and a street outside. Use light gray walls and several neatly arranged bookshelves in the background. The layout must be plausible, furniture proportions accurate, and objects must not intersect or pass through one another. Morning daylight enters through the left window, casting soft shadows on the floor and table. Preserve realistic details of ceramic, glass, metal, wood and fabric. Keep the palette fresh and restrained, mainly white, light wood and green, with even exposure and detail in the shadows. The image should be clear and natural, with the depth and sense of space of a real photograph. Include no people, trademarks, watermarks, decorative borders or readable text, and add nothing unrelated to the cafe.