GPT Image 2.0 · Adobe 4K
English translation. Model names, API fields and executable examples retain their original text, including prompts and comments.
Adobe 4K channel supporting generation and editing. The API model name is always gpt-image-2; explicitly include quality.
The #127 guide has been updated: See the full #127 Image 2 / flare / sunburst API guide for reference counts, dimensions, aspect ratios, examples and actual output differences. The #137 quality-tier notes below are retained and no longer describe #127's current capabilities. Base URLhttps://yyds.seqnode.com/v1 Model gpt-image-2 Authentication Bearer YOUR_API_KEY Output Adobe 4K · medium / high
01 · Model and endpoints
| Capability | Method and endpoint | Content-Type |
|---|---|---|
| Text to image | POST /v1/images/generations |
application/json |
| Image editing | POST /v1/images/edits |
multipart/form-data (recommended) orapplication/json |
The full URLs are https://yyds.seqnode.com/v1/images/generations and https://yyds.seqnode.com/v1/images/edits.
02 · Common request parameters
| Parameter | Required | Description |
|---|---|---|
model |
is the | Fixed value:gpt-image-2 |
prompt |
Yes, for text to image | Generation or editing instructions |
size |
is the | 4K pixel dimensions; commonly3840x2160 or2160x3840 |
quality |
is the | medium orhigh; use the value matching the account's Adobe group |
n |
No | Currently use1 |
response_format |
No | b64_json (recommended) orurl |
output_format |
No | png / jpeg / webp |
image / images |
Yes, for image editing | Repeat multipart uploads usingimage, or use JSONimages[].image_url |
03 · Text to image
JSON request body
{
"model": "gpt-image-2",
"prompt": "高端香水广告,黑色玻璃瓶,金色轮廓光,电影级摄影",
"size": "3840x2160",
"quality": "medium",
"n": 1,
"response_format": "b64_json",
"output_format": "png"
}
cURL
curl https://yyds.seqnode.com/v1/images/generations \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model":"gpt-image-2",
"prompt":"高端香水广告,黑色玻璃瓶,金色轮廓光,电影级摄影",
"size":"3840x2160",
"quality":"medium",
"n":1,
"response_format":"b64_json"
}'
04 · Image editing
Multipart parameters and cURL (local images)
curl https://yyds.seqnode.com/v1/images/edits \
-H "Authorization: Bearer YOUR_API_KEY" \
-F "model=gpt-image-2" \
-F "image=@reference.png" \
-F "prompt=保留主体造型,将背景替换为高级摄影棚,16:9 横图" \
-F "size=3840x2160" \
-F "quality=high" \
-F "n=1" \
-F "response_format=b64_json"
JSON request body (image URLs or Base64)
{
"model": "gpt-image-2",
"images": [
{"image_url": "https://example.com/reference.png"}
],
"prompt": "保留主体,将画面扩展为 9:16 竖版海报",
"size": "2160x3840",
"quality": "high",
"n": 1,
"response_format": "b64_json"
}
JSON cURL
curl https://yyds.seqnode.com/v1/images/edits \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d @edit.json
For multiple references, repeat the multipart -F "image=@FILE" field, or add more images to the JSON images array. Base64 images use data:image/png;base64,BASE64_IMAGE.
05 · Adobe 4K ratios and pixels
| Aspect ratio | Orientation | 4K pixels | quality |
|---|---|---|---|
| 16:9 | Landscape | 3840x2160 |
medium / high |
| 9:16 | Portrait | 2160x3840 |
medium / high |
Selection rule: For the Adobe medium group, send quality="medium"; for Adobe high, send quality="high". The current common 4K routes list 16:9 landscape and 9:16 portrait explicitly. Avoid unconfirmed oversized pixel combinations.
06 · Responses
{
"created": 1700000000,
"data": [{"b64_json": "iVBORw0KGgo..."}],
"model": "gpt-image-2"
}
b64_json returns Base64 images; url returns temporary links.
Last updated: 2026-09-03 · GPT Image 2.0 Adobe 4K