Agnes Image Gen

Agnes AI · Image Generation API

Agnes AI 是全球 Top 10 AI Lab（新加坡），自 2025 年 6 月 1 日起无限期免费开放全模态 API，无需绑定信用卡。

目录

• 快速开始
• 模型选型
• API 参考
• 文生图 Text-to-Image
• 图生图 Image-to-Image
• 多图合成 Multi-Image Composition
• OpenAI SDK 兼容
• Prompt 写作指南
• 错误处理
• 速率限制
• 视频生成

快速开始

1. 获取 API Key

访问 platform.agnes-ai.com 注册，进入设置 → API 密钥 → 创建新密钥。

2. 配置环境变量

export AGNES_API_KEY="sk-your-api-key"

3. 发送第一个请求

curl https://apihub.agnes-ai.com/v1/images/generations \
  -H "Authorization: Bearer $AGNES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agnes-image-2.1-flash",
    "prompt": "A serene mountain lake at golden hour, photorealistic, 8K",
    "size": "1024x768"
  }'

模型选型

选择规则：

API 参考

Base

Base URL:  https://apihub.agnes-ai.com/v1
Endpoint:  POST /images/generations
Auth:      Bearer <API_KEY>
Protocol:  OpenAI Compatible
Price:     Free

请求参数

响应

{
  "created": 1774432125,
  "data": [{ "url": "https://cdn.agnes-ai.com/generated/xxx.png" }],
  "usage": { "generated_images": 1 }
}

文生图 Text-to-Image

模型：agnes-image-2.1-flash

curl

curl https://apihub.agnes-ai.com/v1/images/generations \
  -H "Authorization: Bearer $AGNES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agnes-image-2.1-flash",
    "prompt": "A cyberpunk city street at night, neon signs, rain-slicked pavement, cinematic lighting",
    "size": "1024x768"
  }' | jq -r '.data[0].url'

Python

import httpx, os

resp = httpx.post(
    "https://apihub.agnes-ai.com/v1/images/generations",
    headers={"Authorization": f"Bearer {os.environ['AGNES_API_KEY']}"},
    json={
        "model": "agnes-image-2.1-flash",
        "prompt": "水墨画风格，竹林深处一位侠客独行，晨雾弥漫",
        "size": "1024x768"
    },
    timeout=120
)
image_url = resp.json()["data"][0]["url"]

Node.js

const API_KEY = process.env.AGNES_API_KEY;

const resp = await fetch("https://apihub.agnes-ai.com/v1/images/generations", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${API_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "agnes-image-2.1-flash",
    prompt: "Futuristic city floating in the clouds, golden sunset, 8K",
    size: "1024x768"
  })
});
const { data } = await resp.json();
console.log(data[0].url);

Go

package main

import (
    "bytes"
    "encoding/json"
    "fmt"
    "net/http"
    "os"
)

func main() {
    body, _ := json.Marshal(map[string]any{
        "model":  "agnes-image-2.1-flash",
        "prompt": "A lone samurai under cherry blossoms, ukiyo-e style",
        "size":   "1024x768",
    })
    req, _ := http.NewRequest("POST",
        "https://apihub.agnes-ai.com/v1/images/generations",
        bytes.NewReader(body))
    req.Header.Set("Authorization", "Bearer "+os.Getenv("AGNES_API_KEY"))
    req.Header.Set("Content-Type", "application/json")

    resp, _ := http.DefaultClient.Do(req)
    var result map[string]any
    json.NewDecoder(resp.Body).Decode(&result)
    fmt.Println(result["data"].([]any)[0].(map[string]any)["url"])
}

图生图 Image-to-Image

模型：agnes-image-2.0-flash · 标签：tags: ["img2img"]

Prompt 结构：[改变什么] + keep/preserve [保持什么]

curl

curl https://apihub.agnes-ai.com/v1/images/generations \
  -H "Authorization: Bearer $AGNES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agnes-image-2.0-flash",
    "prompt": "Transform into oil painting with warm tones, keep original composition",
    "size": "1024x768",
    "seed": 42,
    "tags": ["img2img"],
    "extra_body": {
      "image": ["https://example.com/photo.jpg"],
      "response_format": "url"
    }
  }' | jq -r '.data[0].url'

Python

resp = httpx.post(
    "https://apihub.agnes-ai.com/v1/images/generations",
    headers={"Authorization": f"Bearer {os.environ['AGNES_API_KEY']}"},
    json={
        "model": "agnes-image-2.0-flash",
        "prompt": "改为水彩风格，保留原始构图和人物位置",
        "size": "1024x768",
        "seed": 42,
        "tags": ["img2img"],
        "extra_body": {
            "image": ["https://example.com/input.jpg"],
            "response_format": "url"
        }
    },
    timeout=120
)

Node.js

const resp = await fetch("https://apihub.agnes-ai.com/v1/images/generations", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.AGNES_API_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "agnes-image-2.0-flash",
    prompt: "Transform into cyberpunk night scene, keep character position",
    size: "1024x768",
    seed: 42,
    tags: ["img2img"],
    extra_body: {
      image: ["https://example.com/input.jpg"],
      response_format: "url"
    }
  })
});

多图合成 Multi-Image Composition

模型：agnes-image-2.0-flash · 在 extra_body.image 传入多个 URL。

resp = httpx.post(
    "https://apihub.agnes-ai.com/v1/images/generations",
    headers={"Authorization": f"Bearer {os.environ['AGNES_API_KEY']}"},
    json={
        "model": "agnes-image-2.0-flash",
        "prompt": "Fuse these two scenes into one panoramic composition",
        "size": "1024x768",
        "tags": ["img2img"],
        "extra_body": {
            "image": [
                "https://example.com/scene_left.jpg",
                "https://example.com/scene_right.jpg"
            ],
            "response_format": "url"
        }
    },
    timeout=120
)

OpenAI SDK 兼容

Agnes Image API 兼容 OpenAI 的 images.generate 调用协议，可直接替换 base_url：

from openai import OpenAI

client = OpenAI(
    api_key="sk-your-key",
    base_url="https://apihub.agnes-ai.com/v1"
)

# 文生图
resp = client.images.generate(
    model="agnes-image-2.1-flash",
    prompt="赛博朋克城市夜景",
    size="1024x768"
)

# 图生图（需通过 extra_body 传参）
resp = client.images.generate(
    model="agnes-image-2.0-flash",
    prompt="改为水彩风格",
    size="1024x768",
    extra_body={
        "tags": ["img2img"],
        "image": ["https://example.com/input.jpg"],
        "response_format": "url"
    }
)

print(resp.data[0].url)

Prompt 写作指南

文生图公式

主体 + 场景 + 光线 + 风格 + 画质

A [subject] in/at [scene], [lighting], [style], [quality]

示例:
A majestic white tiger standing on a snowy cliff edge,
northern lights in the night sky, rim lighting,
cinematic realism, 8K ultra HD

• 英文效果优于中文（训练数据以英文为主）
• 越具体越好，避免模糊词汇（"beautiful"、"nice"）

图生图公式

改什么 + 保什么

Transform into [new style/effect] while keeping/preserving [elements]

示例:
Transform into anime style with soft pastel colors
while preserving the original composition and character poses

尺寸选择

错误处理

速率限制

免费层有每日调用量和并发限制，对个人开发、原型验证、日常使用完全足够。商业级用量请参考 Agnes AI 官方 的付费方案。

视频生成

Agnes 同时提供视频生成模型 agnes-video-v2.0（异步）：

# 1. 提交生成任务
curl -X POST https://apihub.agnes-ai.com/v1/videos \
  -H "Authorization: Bearer $AGNES_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "agnes-video-v2.0",
    "prompt": "A cat walking on the beach at sunset, gentle waves, 5 seconds",
    "height": 768, "width": 1152,
    "num_frames": 121, "frame_rate": 24
  }'
# → { "id": "task_xxx", "status": "queued" }

# 2. 轮询获取结果
curl https://apihub.agnes-ai.com/v1/videos/task_xxx \
  -H "Authorization: Bearer $AGNES_API_KEY"
# → { "status": "completed", "video_url": "https://..." }

• num_frames 须为 8n+1：81 / 121 / 161 / 241
• frame_rate 推荐 24
• 请求间隔建议 ≥ 10 秒

相关链接