17+ AI Image Models: The Showdown

#ai #imagegeneration #comparison #generativeai

Midjourney v7 · GPT Image 2/1.5/Mini · Nano Banana Pro/2/1 · Kling Kolors v3.0/v2.1 · Seedream 5.0→4.0 · Imagen 4 · Qwen Image · Runway Gen4

Introduction

We ran the same text-to-image prompt through 17 image generation models — all of them available through the useapi.net API — to see how each one interprets it. useapi.net is a third-party API that automates these models across six provider platforms (MiniMax, PixVerse, Dreamina, Kling, Runway, and Google Flow), so every model in this showdown is reachable from one account and token. All images use 16:9 aspect ratio with resolution closest to 1080p — not all models support it since some start at 2K and up.

Google Flow's version of Nano Banana models was too restrictive for this prompt, so Nano Banana generation was done via PixVerse instead.

Why this prompt

This prompt stress-tests NSFW content moderation boundaries, multi-subject composition with precise spatial layout, forced perspective depth of field, contradictory context (bikini + jackhammer), specific animal facial expressions, prop fidelity, single cohesive frame rendering, dynamic action, and clashing tone contrast — all in one image.

Prompt

14 of 17 models used the same prompt:

A visually breathtaking, beautifully absurd, single-frame street photography
shot on a sunny city sidewalk. In the center midground, a gorgeous, stunningly
beautiful young woman with a soft, elegant feminine figure, a flawless glowing
complexion, and a radiant, upbeat laugh is wearing a tiny, elegant black-lace
string bikini. She is effortlessly operating a heavy, rattling, bright yellow
pneumatic jackhammer, busting up the concrete pavement while dust and debris
fly around her beautifully. In the extreme bottom-right foreground of the exact
same scene, standing on the sidewalk and photobombing the camera lens, is a
huge forced-perspective close-up of a city squirrel. The squirrel is holding a
half-eaten hotdog, staring directly into the lens with bulging, mind-blown
"o_O" eyes, completely frozen in comical disbelief. Single cohesive image,
seamless depth of field, no cut-ins, dynamic lighting, crazy situational
comedy, delicate feminine features contrasting with heavy machinery, masterpiece.

Imagen 4 — Google's content moderation required a toned-down version: ...is wearing a stylish summer swimsuit... (all other text identical).

GPT Image 2 — like Imagen 4, OpenAI's content moderation required the toned-down version (...stylish summer swimsuit...).

Midjourney v7 — rewritten for Midjourney's style:

A wide-angle GoPro action-camera shot of a sunny city sidewalk. In the extreme
bottom-right foreground, a shocked squirrel's face photobombs the lens while
holding a hotdog. In the center midground, a stunning woman in a tiny black
lace bikini is clearly visible and in sharp focus, operating a heavy yellow
jackhammer. Concrete debris flying, 8k, deep depth of field, f/11, high-speed
shutter, absurd comedy

How it was generated

All images were generated on the actual provider websites (PixVerse, Dreamina, MiniMax/HailuoAI, Runway, Google Flow, Kling, Midjourney) — our third-party API is just a smart proxy service that automates interaction with the provider websites and does not generate images itself.

Unlimited generations

PixVerse, Google Flow, Dreamina, and Runway support unlimited image generations for select models depending on the underlying website subscription plan.

Introduction

Why this prompt

Prompt

How it was generated

Unlimited generations

Midjourney v7

GPT Image 2

GPT Image 1.5

GPT Image 1 Mini

Nano Banana Pro

Nano Banana 2

Nano Banana

Kling Kolors v3.0

Kling Kolors v2.1

Seedream 5.0 Lite

Seedream 4.6

Seedream 4.5

Seedream 4.1

Seedream 4.0

Imagen 4

Qwen Image

Runway Gen4