OpenAI Images API#

通过 OpenAI 兼容的 Images API 使用 gpt-image-2 生成和编辑图片。

Base URL: https://api.unigateway.ai/v1

文生图#

from openai import OpenAI; import base64
client = OpenAI(api_key="<YOUR_UNIGATEWAY_API_KEY>", base_url="https://api.unigateway.ai/v1")
r = client.images.generate(model="gpt-image-2", prompt="A hero image.")
with open("out.png","wb") as f: f.write(base64.b64decode(r.data[0].b64_json))

curl -sS -X POST "https://api.unigateway.ai/v1/images/generations" \
  -H "Authorization: Bearer $UNIGATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A clean product hero image for an AI gateway dashboard."
  }' > response.json

Python

from openai import OpenAI; import base64
client = OpenAI(api_key="<YOUR_UNIGATEWAY_API_KEY>", base_url="https://api.unigateway.ai/v1")
r = client.images.generate(model="gpt-image-2", prompt="A hero image.")
with open("out.png","wb") as f: f.write(base64.b64decode(r.data[0].b64_json))

TypeScript

import OpenAI from "openai"; import fs from "fs";
const c = new OpenAI({ apiKey: process.env.UNIGATEWAY_API_KEY, baseURL: "https://api.unigateway.ai/v1" });
const r = await c.images.generate({ model: "gpt-image-2", prompt: "A hero image." });
fs.writeFileSync("out.png", Buffer.from(r.data[0].b64_json, "base64"));

多图生成#

curl -sS -X POST "https://api.unigateway.ai/v1/images/generations" \
  -H "Authorization: Bearer $UNIGATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-image-2","prompt":"Four icons: chat, image, video, search.","n":4,"size":"1024x1024","quality":"medium","output_format":"png"}' > batch.json

参数#

流式生成#

启用 stream: true 后，API 以 SSE（Server-Sent Events）逐事件推送，客户端可渐进渲染。

请求#

curl -sS -X POST "https://api.unigateway.ai/v1/images/generations" \
  -H "Authorization: Bearer $UNIGATEWAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-image-2","prompt":"A winter landscape.","stream":true,"partial_images":2}'

SSE 事件格式#

SSE 原始输出示例：

event: image_generation.partial_image
data: {"created_at":1718000000,"type":"image_generation.partial_image","b64_json":"iVBOR...","partial_image_index":0}

event: image_generation.partial_image
data: {"created_at":1718000000,"type":"image_generation.partial_image","b64_json":"iVBOR...","partial_image_index":1}

event: image_generation.completed
data: {"created_at":1718000000,"type":"image_generation.completed","b64_json":"iVBOR..."}

Python 流式消费#

from openai import OpenAI
import base64

client = OpenAI(api_key="<YOUR_UNIGATEWAY_API_KEY>", base_url="https://api.unigateway.ai/v1")

stream = client.images.generate(
    model="gpt-image-2",
    prompt="A winter landscape.",
    stream=True,
    partial_images=2,
)

for event in stream:
    if event.type == "image_generation.partial_image":
        idx = event.partial_image_index
        with open(f"partial_{idx}.png", "wb") as f:
            f.write(base64.b64decode(event.b64_json))
    elif event.type == "image_generation.completed":
        with open("final.png", "wb") as f:
            f.write(base64.b64decode(event.b64_json))

TypeScript 流式消费#

import OpenAI from "openai";
import fs from "fs";

const client = new OpenAI({ apiKey: process.env.UNIGATEWAY_API_KEY, baseURL: "https://api.unigateway.ai/v1" });

const stream = await client.images.generate({
  model: "gpt-image-2",
  prompt: "A winter landscape.",
  stream: true,
  partial_images: 2,
});

for await (const event of stream) {
  if (event.type === "image_generation.partial_image") {
    const idx = event.partial_image_index;
    fs.writeFileSync(`partial_${idx}.png`, Buffer.from(event.b64_json, "base64"));
  } else if (event.type === "image_generation.completed") {
    fs.writeFileSync("final.png", Buffer.from(event.b64_json, "base64"));
  }
}

端点兼容性#

不同上游端点对流式生成的支持程度不同。仅 OpenAI 官方和 Azure OpenAI 支持真正的 SSE 流式推送（含渐进 partial_image 事件）。其他端点虽然接受 stream:true，但可能在图片完全生成后才一次性返回 JSON，无法实现渐进渲染。

注意：如需真正的流式体验（渐进渲染），请将 gpt-image-2 流式请求路由到支持 SSE 的端点（OpenAI 官方或 Azure OpenAI）。

保存返回#

jq -r '.data[0].b64_json' response.json | base64 -D > output.png    # macOS
jq -r '.data[0].b64_json' response.json | base64 --decode > output.png  # Linux

编辑 / 合成 / 重绘#

均使用 POST /v1/images/edits，multipart/form-data。

单图编辑：

curl -sS -X POST "https://api.unigateway.ai/v1/images/edits" \
  -H "Authorization: Bearer $UNIGATEWAY_API_KEY" \
  -F "model=gpt-image-2" -F "image[]=@room.png" \
  -F "prompt=将沙发改为米白色，其余保持不变。" \
  -F "quality=high" -F "size=1024x1024" -F "output_format=png" > edit.json

多参考图合成：

curl -sS -X POST "https://api.unigateway.ai/v1/images/edits" \
  -H "Authorization: Bearer $UNIGATEWAY_API_KEY" \
  -F "model=gpt-image-2" \
  -F "image[]=@item1.png" -F "image[]=@item2.png" -F "image[]=@item3.png" \
  -F "prompt=将全部物品合成一张白底产品照。" \
  -F "quality=high" -F "output_format=png" > composite.json

局部重绘：

curl -sS -X POST "https://api.unigateway.ai/v1/images/edits" \
  -H "Authorization: Bearer $UNIGATEWAY_API_KEY" \
  -F "model=gpt-image-2" -F "mask=@mask.png" -F "image[]=@src.png" \
  -F "prompt=在 mask 区域画一只粉色火烈鸟。" > inpaint.json

Mask 要求：尺寸一致、格式相同、≤ 50 MB、含 alpha 通道。

响应格式#

{ "created": 1710000000, "data": [{ "b64_json": "..." }] }

常见错误#