nano-banana-2.1 图片生成 API

快速信息

获取 API Key

登录 DeepSeek Tavern 控制台，在「令牌管理」中创建 API Key。详细步骤参考：API 密钥获取指南。

Base URL 统一为 https://deepseektavern.com/v1，认证方式：Authorization: Bearer <你的API密钥>。下文示例中的 YOUR_API_KEY 请替换为你自己的 Key。

方式一：OpenAI 图片生成端点

POST https://deepseektavern.com/v1/images/generations

curl 示例：文生图

curl https://deepseektavern.com/v1/images/generations \
  --max-time 300 \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "nano-banana-2.1",
    "prompt": "一位白发少女站在樱花树下，新海诚电影风格，柔和暖色调",
    "aspect_ratio": "16:9",
    "image_size": "2K",
    "n": 1
  }'

curl 示例：图生图（参考图）

curl https://deepseektavern.com/v1/images/generations \
  --max-time 300 \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "nano-banana-2.1",
    "prompt": "保持人物外貌不变，把背景换成雪夜的古风街道",
    "images": ["https://example.com/my-character.png"],
    "size": "1024x1536"
  }'

Python 示例（OpenAI SDK）

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://deepseektavern.com/v1",
    timeout=300,  # 生成一张约 1-2 分钟，超时要设到 180 秒以上
)

resp = client.images.generate(
    model="nano-banana-2.1",
    prompt="一只在星空下弹吉他的猫，吉卜力风格",
    size="1024x1024",
    n=1,
    # SDK 未内置的参数通过 extra_body 传入
    extra_body={
        "image_size": "2K",
        # "aspect_ratio": "16:9",
        # "images": ["https://example.com/ref.png"],
    },
)

for img in resp.data:
    print(img.url)  # 结果链接请及时下载保存

Python 示例：保存 b64_json 结果

import base64
from openai import OpenAI

client = OpenAI(api_key="YOUR_API_KEY", base_url="https://deepseektavern.com/v1", timeout=300)

resp = client.images.generate(
    model="nano-banana-2.1",
    prompt="赛博朋克风格的夜市街道，霓虹灯，雨后反光",
    response_format="b64_json",
    extra_body={"aspect_ratio": "9:16"},
)
with open("result.png", "wb") as f:
    f.write(base64.b64decode(resp.data[0].b64_json))

方式二：改图端点

POST https://deepseektavern.com/v1/images/edits，支持两种请求格式：

• multipart/form-data：字段 image 上传图片文件，其余参数（model、prompt 等）作为表单字段
• JSON：参考图放在 images 数组里（URL 或 data URI），参数与生成端点相同

curl 示例：multipart 上传文件

curl https://deepseektavern.com/v1/images/edits \
  --max-time 300 \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F model=nano-banana-2.1 \
  -F image=@input.png \
  -F prompt="把衣服换成黑色哥特风长裙，其余保持不变"

curl 示例：JSON 格式

curl https://deepseektavern.com/v1/images/edits \
  --max-time 300 \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "nano-banana-2.1",
    "prompt": "给人物加上一对白色羽翼",
    "images": ["https://example.com/input.png"]
  }'

Python 示例（OpenAI SDK）

from openai import OpenAI

client = OpenAI(api_key="YOUR_API_KEY", base_url="https://deepseektavern.com/v1", timeout=300)

with open("input.png", "rb") as f:
    resp = client.images.edit(
        model="nano-banana-2.1",
        image=f,
        prompt="把背景换成海边日落",
    )
print(resp.data[0].url)

方式三：Chat 端点（适合 SillyTavern 等聊天客户端）

POST https://deepseektavern.com/v1/chat/completions，在 messages 里写提示词即可；需要参考图时，按 OpenAI 多模态格式附 image_url。返回内容是 markdown 图片（形如 ![image](https://...)），聊天客户端可直接渲染。

curl 示例

curl https://deepseektavern.com/v1/chat/completions \
  --max-time 300 \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "nano-banana-2.1",
    "messages": [
      {"role": "user", "content": "画一只在星空下弹吉他的猫，吉卜力风格，16:9"}
    ]
  }'

Python 示例：附参考图

from openai import OpenAI

client = OpenAI(api_key="YOUR_API_KEY", base_url="https://deepseektavern.com/v1", timeout=300)

resp = client.chat.completions.create(
    model="nano-banana-2.1",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "参考这张图的人物，画她在图书馆看书的场景"},
            {"type": "image_url", "image_url": {"url": "https://example.com/ref.png"}},
        ],
    }],
)
# 返回 markdown 图片，例如 ![image](https://...)
print(resp.choices[0].message.content)

在 SillyTavern 中使用

1. API 类型选择 Chat Completion
2. Chat Completion Source 选择 Custom (OpenAI-compatible)
3. 填入端点：https://deepseektavern.com/v1
4. 填入你的 API Key
5. 模型名称手动输入：nano-banana-2.1

方式四：Gemini 原生端点

POST https://deepseektavern.com/v1beta/models/nano-banana-2.1:generateContent

curl 示例

curl "https://deepseektavern.com/v1beta/models/nano-banana-2.1:generateContent" \
  --max-time 300 \
  -H "Content-Type: application/json" \
  -H "x-goog-api-key: YOUR_API_KEY" \
  -d '{
    "contents": [{"parts": [{"text": "水墨画风格的江南小镇，细雨，乌篷船"}]}],
    "generationConfig": {"imageConfig": {"aspectRatio": "3:2", "imageSize": "1K"}}
  }'

Python 示例（requests）

import base64
import requests

API_KEY = "YOUR_API_KEY"
url = "https://deepseektavern.com/v1beta/models/nano-banana-2.1:generateContent"

with open("ref.png", "rb") as f:
    ref_b64 = base64.b64encode(f.read()).decode()

body = {
    "contents": [{
        "parts": [
            {"text": "参考这张图的人物，画成 Q 版贴纸风格"},
            {"inlineData": {"mimeType": "image/png", "data": ref_b64}},
        ]
    }],
    "generationConfig": {"imageConfig": {"aspectRatio": "1:1", "imageSize": "1K"}},
}
r = requests.post(url, json=body, headers={"x-goog-api-key": API_KEY}, timeout=300)
r.raise_for_status()

for i, part in enumerate(r.json()["candidates"][0]["content"]["parts"]):
    if "inlineData" in part:
        with open(f"out_{i}.png", "wb") as f:
            f.write(base64.b64decode(part["inlineData"]["data"]))

计费说明