文生图

← 生图模型目录 · 接口与鉴权先按快速开始设置 NONELINEAR_API_KEY；Python 需安装 openai。每个代码块可独立使用，只发送一次生成请求。图片地址请替换为您自己的公网 HTTPS 图片链接。

Python SDK 的扩展参数放在 extra_body 中；cURL 把这些字段放在 JSON 顶层。示例中的模型 ID 和参数需成套使用，不要只替换模型而保留不适用的尺寸。

用 aspect_ratio 控制比例：适用于 Nano Banana、FLUX、MiniMax 等，以 gemini-2.5-flash-image 为例。

curl -sS --max-time 300 \
  https://api.nonelinear.com.cn/v1/images/generations \
  -H "Authorization: Bearer $NONELINEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-2.5-flash-image",
    "prompt": "A cinematic wide-angle photo of a futuristic city skyline at sunset, rich details, natural lighting",
    "response_format": "url",
    "aspect_ratio": "16:9"
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.nonelinear.com.cn/v1",
    api_key=os.environ["NONELINEAR_API_KEY"],
    timeout=300,
    max_retries=0,
)

response = client.images.generate(
    model="gemini-2.5-flash-image",
    prompt="A cinematic wide-angle photo of a futuristic city skyline at sunset, rich details, natural lighting",
    response_format="url",
    extra_body={"aspect_ratio": "16:9"},
)
for image in response.data or []:
    print(image.url or image.b64_json)

同时用 aspect_ratio 和 size 控制比例与清晰度：适用于 Nano Banana 2 / Pro、Kling、Vidu 等，以 gemini-3-pro-image-preview 为例。

curl -sS --max-time 300 \
  https://api.nonelinear.com.cn/v1/images/generations \
  -H "Authorization: Bearer $NONELINEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-pro-image-preview",
    "prompt": "A high fidelity studio portrait of a glass perfume bottle on black acrylic, sharp reflections, commercial photography",
    "response_format": "url",
    "aspect_ratio": "4:3",
    "size": "2K"
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.nonelinear.com.cn/v1",
    api_key=os.environ["NONELINEAR_API_KEY"],
    timeout=300,
    max_retries=0,
)

response = client.images.generate(
    model="gemini-3-pro-image-preview",
    prompt="A high fidelity studio portrait of a glass perfume bottle on black acrylic, sharp reflections, commercial photography",
    response_format="url",
    size="2K",
    extra_body={"aspect_ratio": "4:3"},
)
for image in response.data or []:
    print(image.url or image.b64_json)

使用 gemini-3.1-flash-image-preview（Nano Banana 2）时写法相同，例如 "aspect_ratio": "1:1"、"size": "1K"。 只用 size 控制尺寸：适用于 Seedream、Qwen、Wan、GPT Image 等，以 doubao-seedream-4-0-250828 为例。

curl -sS --max-time 300 \
  https://api.nonelinear.com.cn/v1/images/generations \
  -H "Authorization: Bearer $NONELINEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-4-0-250828",
    "prompt": "星际穿越，黑洞里冲出一辆支离破碎的复古列车，电影大片质感，强烈视觉冲击力，深蓝色调，光线追踪，景深，夸张的广角透视",
    "response_format": "url",
    "size": "2K",
    "watermark": false
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.nonelinear.com.cn/v1",
    api_key=os.environ["NONELINEAR_API_KEY"],
    timeout=300,
    max_retries=0,
)

response = client.images.generate(
    model="doubao-seedream-4-0-250828",
    prompt="星际穿越，黑洞里冲出一辆支离破碎的复古列车，电影大片质感，强烈视觉冲击力，深蓝色调，光线追踪，景深，夸张的广角透视",
    response_format="url",
    size="2K",
    extra_body={"watermark": False},
)
for image in response.data or []:
    print(image.url or image.b64_json)

Seedream 5.0 Pro：可额外指定输出格式和提示词优化模式。追求速度时，把 mode 改为 fast、output_format 改为 jpeg。

curl -sS --max-time 600 \
  https://api.nonelinear.com.cn/v1/images/generations \
  -H "Authorization: Bearer $NONELINEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-5-0-pro-260628",
    "prompt": "一只透明玻璃茶壶放在深色胡桃木桌面上，壶中有琥珀色茶汤，左侧柔和晨光，背景为简洁的现代厨房，1:1构图，真实产品摄影，细节清晰",
    "response_format": "url",
    "size": "1K",
    "output_format": "png",
    "watermark": false,
    "optimize_prompt_options": {
      "mode": "standard"
    }
  }'

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.nonelinear.com.cn/v1",
    api_key=os.environ["NONELINEAR_API_KEY"],
    timeout=300,
    max_retries=0,
)

response = client.with_options(timeout=600).images.generate(
    model="doubao-seedream-5-0-pro-260628",
    prompt="一只透明玻璃茶壶放在深色胡桃木桌面上，壶中有琥珀色茶汤，左侧柔和晨光，背景为简洁的现代厨房，1:1构图，真实产品摄影，细节清晰",
    response_format="url",
    size="1K",
    extra_body={
        "output_format": "png",
        "watermark": False,
        "optimize_prompt_options": {"mode": "standard"},
    },
)
for image in response.data or []:
    print(image.url or image.b64_json)

检查结果

以 data 中存在 url 或 b64_json 判断出图，并及时保存。未出图、参考图不可访问或超时，见生图排障。

开发基础

生图任务

框架集成

检查结果

​检查结果

检查结果