01
引言
OpenAI的图像生成器从未跻身最佳AI模型之列。一直以来，Midjourney、Flux和Imagen才是这个领域的佼佼者。似乎OpenAI过于专注于文本生成，以至于忽略了对其图像模型的升级。
然而，就在几天前，OpenAI推出了一款全新的GPT-4o模型，内置了图像生成功能，直接将竞争对手远远甩在身后。这一整合使得该模型能够直接从文本提示中生成精确、逼真的图像，同时利用其内置知识和聊天上下文来提升图像质量、文本准确性、角色一致性和提示连贯性。
目前有数款图像生成器正在角逐行业巅峰：
- Midjourney V6
- Black Forest Labs 推出的 Flux 1.1 Pro Ultra
- OpenAI 的 GPT-4o
为了帮助大家选择最适合的工具，我用相同的提示词让这三款AI模型生成图像，并将结果并排对比展示。
闲话少说，我们直接开始吧！
02
第一个提示词如下：
Prompt: A majestic lion kneeling near a crystal-clear water source in the savannah. It drinks slowly, its eyes fixed on the clear water, while its golden mane gently sways in the breeze. The water reflects its image, creating a perfect symmetry between the lion and its reflection.
The grass around the water source is fresh and lush, contrasting with the warmth of the savannah. The wild landscape stretches in the background, slightly blurred, highlighting the tranquility of this serene moment
对比结果如下：
三幅作品的视觉效果都堪称震撼，狮子的存在感更是气势逼人。但平心而论，Midjourney在此次对决中展现出了惊人的细节把控力——画面质感更为柔和自然，仿佛国家地理杂志的专业摄影作品。背景的柔焦虚化与水面倒影的平滑过渡，让整幅画面既真实又充满静谧之美。这张作品完美还原了文字描述中狮子在晨曦湖畔饮水的静谧意境，因此成为大家最钟爱的一幅作品。
第二个提示词如下：
Prompt: A vintage logo design featuring the brand name “Golden Roots“ in a retro serif font. The logo includes ornate details like vines and leaves, with muted earth tones. The design has a hand-drawn, artisanal feel.
第三个提示词如下：
Prompt: Boho-style interior captured in perspective, featuring a soft light beige sofa. Above the sofa, a gallery of same-sized paintings hangs on the wall, each framed in thin light wooden frames with completely white canvases inside.
The sofa is adorned with brown cushions and a light throw. In front of the sofa, there is a light wooden coffee table. A boho-style floor lamp stands nearby. Soft warm-white sunlight filters through the window, casting gentle shadows across the scene
Prompt: A middle-aged man wearing a transparent spherical helmet filled with water, with small orange goldfish swimming inside. The helmet has a breathing apparatus attached to the mouth and a snorkel-style valve.
The man wears clear swimming goggles inside the helmet. His face appears slightly distorted by the pressure and curvature of the helmet. He is standing in a crowd of people wearing winter clothes, under bright daylight, outdoor protest setting, hyperrealistic, photojournalistic style
03
-
能将代码渲染成图像 -
支持在图像上涂鸦指令 -
可生成透明背景图像
以下是一些示例场景：
它能根据代码生成用户界面。用户只需在ChatGPT中上传或编写代码，即可要求AI将代码输入转换为可视化图像。
GPT-4o rendering an image from code
对于开发者和网页设计师而言，这可能会带来革命性的改变。目前没有任何竞争对手能够实现这一功能。
这款全新的图像模型还能精准处理包含多个对象的复杂指令。来看看EP团队提供的这个示例：
editing an image from instructions written on an image
GPT-4o处理图像中嵌入文本指令的能力实在令人惊叹！大家只需在图像上随手涂鸦标注想要的改动，AI就会自动帮你完成剩下的工作，这种智能程度简直疯狂！
Prompt: Turn this image into a cute sticker with transparent background
这意味着相关从业者无需再将照片导出到Photoshop手动抠图去背景，这简直是效率革命的福音。
最后不得不提，GPT-4o在风格迁移方面堪称惊艳。眼下全网正被吉卜力工作室风格图像刷屏，你只需上传照片，就能让AI将其重塑成宫崎骏动画般的艺术风格。
欢迎大家持续关注！
一起学习，共同进步！！！点击上方小卡片关注我
添加个人微信，进专属粉丝群！
