跳转到主要内容
POST
/
v1
/
images
/
generations
curl -s https://api.cometapi.com/v1/images/generations \
  -H "Authorization: Bearer $COMETAPI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-5-0-260128",
    "prompt": "a cat wearing a scarf sitting by the window watching rain, illustration style",
    "size": "2K",
    "response_format": "url",
    "watermark": false
  }' | jq -r '.data[].url'
{
  "created": 123,
  "data": [
    {
      "url": "<string>",
      "b64_json": "<string>",
      "size": "<string>"
    }
  ],
  "usage": {
    "generated_images": 123,
    "output_tokens": 123,
    "total_tokens": 123
  },
  "model": "<string>"
}

Documentation Index

Fetch the complete documentation index at: https://apidoc.cometapi.com/llms.txt

Use this file to discover all available pages before exploring further.

​
概述

通过标准的 OpenAI 兼容 /v1/images/generations 端点，使用 ByteDance 的 Seedream 模型生成图像。支持 文生图 和 图生图 两种生成方式。 完整模型列表请参见模型页面。如需查看提供方的详细参数文档，请参见 VolcEngine documentation。
使用顺序（多图）生成时，将 sequential_image_generation 设置为 auto，并可选择配置 sequential_image_generation_options.max_images 来限制输出图像数量。参考图像总数与生成图像总数之和必须 ≤ 15。

​
图生图

通过 image 参数（URL 数组或 base64 data URI）传入一张或多张参考图像，并附带文本提示词。最多支持 14 张参考图像。

​
返回的 URL 出现 AccessDenied？

返回的 url 是一个带签名的对象存储链接，其中包含 & 和其他特殊字符。
如果你复制原始 JSON 并直接粘贴到浏览器或脚本中，shell 转义或终端截断可能会去掉签名参数，从而导致：
\{"Code":"AccessDenied","Message":"Access Denied","EC":"0003-00000015"\}
如何修复：
  • curl：通过 | jq -r '.data[].url' 管道提取 URL，避免污染。
  • Python：使用 result.data[0].url 或 image["url"] —— 该字符串本身已经完整；直接传给 requests.get() 或 webbrowser.open() 即可。
  • JavaScript：同理，使用 result.data[0].url。
  • 通用：确保 URL 中的 & 和 % 没有被截断或二次转义。

授权

​
Authorization
string
header
必填

Bearer token authentication. Use your CometAPI key.

请求体

application/json

Parameters for Seedream image generation (text-to-image and image-to-image).

​
model
string
默认值:doubao-seedream-5-0-260128
必填

Model to use. See the Models page for current Seedream model IDs.

​
prompt
string
默认值:a cat wearing a scarf sitting by the window
必填

Text prompt describing the image to generate. Supports both Chinese and English. Recommended max 300 Chinese characters or 1000 English characters.

​
response_format
enum<string>
默认值:url

Output format. url returns a temporary image link (valid for 24 hours); b64_json returns base64-encoded image data.

可用选项:
url,
b64_json
​
size
string
默认值:2048x2048

Output resolution. Presets 2K, 4K, or exact pixels like 2048x2048, 2304x1728, 2848x1600, 1600x2848, etc. Default: 2048x2048.

​
image
string[]

Reference image(s) for image-to-image generation. Each item is a public URL or base64 data URI (data:image/png;base64,...). Supports up to 14 reference images. Supported formats: jpeg, png, webp, bmp, tiff, gif.

​
watermark
boolean
默认值:true

When true, adds an 'AI Generated' watermark to the bottom-right corner of the output image.

​
guidance_scale
number
默认值:2.5

Controls how closely the output follows the prompt. Higher values increase adherence but may reduce naturalness. Range: 1.0–10.0. Not supported by Seedream 5.0 (ignored).

​
seed
integer
默认值:-1

Random seed for reproducible outputs. Range: -1 to 2147483647. Use -1 (default) for random generation. Legacy parameter — ignored by newer models.

​
sequential_image_generation
enum<string>
默认值:disabled

Controls sequential (multi-image) generation mode.

  • auto: model decides whether to generate a sequence based on the prompt
  • disabled: always generate a single image
可用选项:
auto,
disabled
​
sequential_image_generation_options
object

Configuration for sequential (multi-image) generation. Only effective when sequential_image_generation is auto.

​
output_format
enum<string>
默认值:jpeg

Output image file format. Options: png, jpeg.

可用选项:
png,
jpeg

响应

200 - application/json

Successful Response

​
created
integer
必填

Unix timestamp (seconds) when the image was generated.

​
data
object[]
必填

Array of generated image objects.

​
usage
object
必填

Resource usage information for this request.

​
model
string

The model used for generation.