用 Gemini 生成图像

curl https://api.pipellm.ai/v1beta/models/gemini-3-pro-image:generateContent \
  -H "x-goog-api-key: $PIPELLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "role": "user",
      "parts": [{"text": "A red bicycle leaning against a cafe window, morning light"}]
    }]
  }'

import os
from google import genai

client = genai.Client(
    api_key=os.environ["PIPELLM_API_KEY"],
    http_options={"base_url": "https://api.pipellm.ai"},
)

response = client.models.generate_content(
    model="gemini-3-pro-image",
    contents="A red bicycle leaning against a cafe window, morning light",
)

for part in response.parts:
    if part.inline_data is not None:
        part.as_image().save("bicycle.png")

import fs from "node:fs";
import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({
  apiKey: process.env.PIPELLM_API_KEY,
  httpOptions: { baseUrl: "https://api.pipellm.ai" },
});

const response = await ai.models.generateContent({
  model: "gemini-3-pro-image",
  contents: "A red bicycle leaning against a cafe window, morning light",
});

for (const part of response.candidates[0].content.parts) {
  if (part.inlineData) {
    fs.writeFileSync("bicycle.png", Buffer.from(part.inlineData.data, "base64"));
  }
}

{
  "candidates": [{
    "content": {
      "role": "model",
      "parts": [
        { "text": "Here is the bicycle scene you asked for." },
        {
          "inlineData": {
            "mimeType": "image/png",
            "data": "iVBORw0KGgoAAAANSUhEUgAA..."
          }
        }
      ]
    },
    "finishReason": "STOP"
  }],
  "usageMetadata": {
    "promptTokenCount": 12,
    "candidatesTokenCount": 1290,
    "totalTokenCount": 1302
  }
}

POST

v1beta

models

{model}

:generateContent

curl https://api.pipellm.ai/v1beta/models/gemini-3-pro-image:generateContent \
  -H "x-goog-api-key: $PIPELLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "role": "user",
      "parts": [{"text": "A red bicycle leaning against a cafe window, morning light"}]
    }]
  }'

import os
from google import genai

client = genai.Client(
    api_key=os.environ["PIPELLM_API_KEY"],
    http_options={"base_url": "https://api.pipellm.ai"},
)

response = client.models.generate_content(
    model="gemini-3-pro-image",
    contents="A red bicycle leaning against a cafe window, morning light",
)

for part in response.parts:
    if part.inline_data is not None:
        part.as_image().save("bicycle.png")

import fs from "node:fs";
import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({
  apiKey: process.env.PIPELLM_API_KEY,
  httpOptions: { baseUrl: "https://api.pipellm.ai" },
});

const response = await ai.models.generateContent({
  model: "gemini-3-pro-image",
  contents: "A red bicycle leaning against a cafe window, morning light",
});

for (const part of response.candidates[0].content.parts) {
  if (part.inlineData) {
    fs.writeFileSync("bicycle.png", Buffer.from(part.inlineData.data, "base64"));
  }
}

{
  "candidates": [{
    "content": {
      "role": "model",
      "parts": [
        { "text": "Here is the bicycle scene you asked for." },
        {
          "inlineData": {
            "mimeType": "image/png",
            "data": "iVBORw0KGgoAAAANSUhEUgAA..."
          }
        }
      ]
    },
    "finishReason": "STOP"
  }],
  "usageMetadata": {
    "promptTokenCount": 12,
    "candidatesTokenCount": 1290,
    "totalTokenCount": 1302
  }
}

Gemini 图像模型不走 /v1/images/*。它和文本模型一样通过 generateContent 生成图像，图像字节以 inline part 的形式，和模型输出的文本一起返回。 GPT 图像模型请改用 POST /v1/images/generations。完整的文生图、图生图示例见 Nano Banana。

模型 ID 会随时间变化。用 GET /v1/models 查询你的账户当前可用的模型。

端点

模式	路径
非流式	`POST /v1beta/models/{model}:generateContent`
流式	`POST /v1beta/models/{model}:streamGenerateContent`

和所有 Gemini 路由一样，用 x-goog-api-key 认证。见 Gemini 总览。

array

必填

对话轮次。文生图只需一个 user 轮次、一个文本 part；图生图则在同一轮次里追加源图的 inline part。

显示 contents[]

string

user 或 model。

array

必填

文本 part 和图像 inline part 的混合。

显示 parts[]

string

提示词文本。

object

源图，由 mimeType（例如 image/png）和 base64 的 data 组成。

object

Gemini 标准生成配置。图像模型还接受尺寸相关的控制项，但各模型支持的宽高比和分辨率不同，依赖具体取值前请先确认该模型的能力。

curl https://api.pipellm.ai/v1beta/models/gemini-3-pro-image:generateContent \
  -H "x-goog-api-key: $PIPELLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{
      "role": "user",
      "parts": [{"text": "A red bicycle leaning against a cafe window, morning light"}]
    }]
  }'

import os
from google import genai

client = genai.Client(
    api_key=os.environ["PIPELLM_API_KEY"],
    http_options={"base_url": "https://api.pipellm.ai"},
)

response = client.models.generate_content(
    model="gemini-3-pro-image",
    contents="A red bicycle leaning against a cafe window, morning light",
)

for part in response.parts:
    if part.inline_data is not None:
        part.as_image().save("bicycle.png")

import fs from "node:fs";
import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({
  apiKey: process.env.PIPELLM_API_KEY,
  httpOptions: { baseUrl: "https://api.pipellm.ai" },
});

const response = await ai.models.generateContent({
  model: "gemini-3-pro-image",
  contents: "A red bicycle leaning against a cafe window, morning light",
});

for (const part of response.candidates[0].content.parts) {
  if (part.inlineData) {
    fs.writeFileSync("bicycle.png", Buffer.from(part.inlineData.data, "base64"));
  }
}

{
  "candidates": [{
    "content": {
      "role": "model",
      "parts": [
        { "text": "Here is the bicycle scene you asked for." },
        {
          "inlineData": {
            "mimeType": "image/png",
            "data": "iVBORw0KGgoAAAANSUhEUgAA..."
          }
        }
      ]
    },
    "finishReason": "STOP"
  }],
  "usageMetadata": {
    "promptTokenCount": 12,
    "candidatesTokenCount": 1290,
    "totalTokenCount": 1302
  }
}

array

生成结果。

显示 candidates[]

array

交错排列的输出。每个 part 要么是 text，要么是 inlineData，不会同时有。一次响应里两种可以各有多个，所以要遍历全部 part，不要只读 parts[0]。

string

例如 image/png。

string

Base64 编码的图像字节。

string

成功时为 STOP。因安全策略中止时不会返回图像 part。

object

Token 用量。生成的图像按图像输出 token 计费，这部分是费用的主要来源，见价格。

解析响应

两个习惯能省掉不少麻烦：

遍历每一个 part。 让模型生成带插图的说明时，文本和图像 part 是交错的。只取第一个 part 会静默丢内容。
要能接受纯文本响应。 提示词因安全策略被拒时，响应仍是 200，只有文本 part，没有 inlineData。解码前先确认图像 part 存在。

Gemini 生成的图像带有 SynthID 水印。

错误

Gemini 路由返回 Gemini 的错误结构，不是 OpenAI 那套。完整映射见错误处理。

400 INVALID_ARGUMENT

contents 结构有误、inline part 的 mimeType 不受支持，或尺寸取值模型不接受。

401 UNAUTHENTICATED

缺少或填错 x-goog-api-key。

429 RESOURCE_EXHAUSTED

见速率限制。

API 指南

API 参考

用 Gemini 生成图像

端点

解析响应

错误

​端点

​解析响应

​错误

端点

解析响应

错误