GPT-5.4 Image-2 API文档

概述

GPT-5.4 Image-2 是一款多模态图像模型，支持：

• 文生图（Text-to-Image）
• 图生图（Image-to-Image）
• 图片编辑（Edits / Inpainting）
• 多图融合（Multi-image Composition）
• 图片理解（Vision）
• Base64 图像输出

完全兼容 OpenAI API 风格。

Base URL

https://api.yisu.com

认证

Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

统一模型

gpt-5.4-image-2

1. Images Generations

Endpoint

POST /v1/images/generations

请求参数

size

1024x1024
1024x1536
1536x1024
auto

quality

low
medium
high
auto

示例

curl https://api.yisu.com/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "model": "gpt-5.4-image-2",
  "prompt": "赛博朋克城市夜景",
  "size": "1024x1024",
  "n": 1,
  "quality": "high",
  "response_format": "b64_json"
}'

响应

{
  "created": 1779778659,
  "data": [
    {
      "index": 0,
      "b64_json": "iVBORw0KGgoAAAANSUhEUgAA..."
    }
  ]
}

2. Images Edits

Endpoint

POST /v1/images/edits

请求方式

multipart/form-data

参数

mask 规则

示例

curl https://api.yisu.com/v1/images/edits \
-H "Authorization: Bearer YOUR_API_KEY" \
-F "model=gpt-5.4-image-2" \
-F "prompt=给人物添加墨镜并增强光影" \
-F "image=@input.png" \
-F "response_format=b64_json"

响应

{
  "created": 1779778659,
  "data": [
    {
      "index": 0,
      "b64_json": "iVBORw0KGgoAAAANSUhEUgAA..."
    }
  ]
}

3. 图生图 / 多图融合（Chat Completions）

Endpoint

POST /v1/chat/completions

请求示例

{
  "model": "gpt-5.4-image-2",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "把人物背景改成雪山"
        },
        {
          "type": "image",
          "image": "data:image/png;base64,IMAGE"
        }
      ]
    }
  ]
}

响应

{
  "id": "chatcmpl_xxx",
  "object": "chat.completion",
  "created": 1779778659,
  "model": "gpt-5.4-image-2",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "",
        "images": [
          {
            "type": "image_url",
            "image_url": {
              "url": "data:image/png;base64,..."
            }
          }
        ]
      },
      "finish_reason": "stop"
    }
  ]
}

4. 图片理解（Vision）

请求

{
  "model": "gpt-5.4-image-2",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "描述这张图片"
        },
        {
          "type": "image",
          "image": "data:image/jpeg;base64,..."
        }
      ]
    }
  ]
}

响应

{
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "图片中是一只猫坐在沙发上。",
        "images": []
      }
    }
  ]
}

5. 图片返回规则（统一）

生成图片位置

Chat API：
choices[0].message.images[]
Images API：
data[].b64_json

图片格式

data:image/png;base64,...
data:image/jpeg;base64,...

特点

• 不返回公网 URL
• 仅返回 Base64
• 客户端需自行存储

6. Base64 解码示例

Python

import base64
img = response["data"][0]["b64_json"]
with open("out.png", "wb") as f:
    f.write(base64.b64decode(img))

PHP

$img = $response['data'][0]['b64_json'];
file_put_contents("out.png", base64_decode($img));

7. 流式输出（Chat）

SSE 示例

data: {
  "choices": [
    {
      "delta": {
        "content": "正在生成..."
      }
    }
  ]
}
data: {
  "choices": [
    {
      "delta": {
        "images": [
          {
            "image_url": {
              "url": "data:image/png;base64,..."
            }
          }
        ]
      }
    }
  ]
}
data: [DONE]

8. 错误响应

API Key 无效

{
  "error": {
    "message": "Invalid API key.",
    "type": "authentication_error",
    "code": "invalid_api_key"
  }
}

参数错误

{
  "error": {
    "message": "Invalid request.",
    "type": "invalid_request_error",
    "code": "invalid_request"
  }
}

图片错误

{
  "error": {
    "message": "Invalid image format.",
    "type": "invalid_request_error",
    "code": "invalid_image"
  }
}

9. 支持能力总结

• 文生图
• 图生图
• 图片编辑
• 局部重绘
• 多图融合
• 图片理解
• Base64 输出