### Which requests currently use upscaling? **Requests pass through the upscaling proxy only when routed to the local upscaling account. Size-based automatic routing is currently disabled, so requesting 2K/4K or nonstandard dimensions in \#75 does not guarantee that the proxy is used.** \#77 and \#127 are not bound to this local upscaling account. Inside the proxy, the upstream source size follows the target orientation. The result is scaled proportionally and center-cropped to the target dimensions: | Target orientation | Upstream source size | Target examples | |--------------------|----------------------|--------------------------| | Square | `1024x1024` | `2048x2048`, `4096x4096` | | Landscape | `1536x1024` | `2048x1152`, `3840x2160` | | Portrait | `1024x1536` | `1152x2048`, `2160x3840` | Here, upscaling means image resampling and cropping, not an additional AI upscaling model. Smaller targets may be downscaled; cropping may remove edge content.

### Standard reference dimensions The following 14 dimensions come from the site's routing configuration template. **That template is currently disabled**. They explain “standard / nonstandard” dimensions and do not describe live automatic routing. | Orientation | Standard reference sizes | |----|----| | Square | `1024x1024`, `1280x1280` | | Landscape | `1536x1024`, `1536x1152`, `1360x1024`, `1536x864`, `1600x900`, `1824x1024` | | Portrait | `1024x1536`, `1152x1536`, `1024x1360`, `864x1536`, `900x1600`, `1024x1824` |

### Ratios and nonstandard size examples Common ratios: **1:1、16:9、9:16、4:3、3:4、3:2、2:3、3:1、1:3、21:9、9:21、5:4、4:5**. The dimensions below are outside the standard reference set above. They are request templates, **not an individually verified allowlist for all models or a billing-tier table**. | Aspect ratio | Nonstandard`size` example | |--------------|----------------------------------------------------| | 1:1 | `2048x2048`, `2880x2880`, `4096x4096` | | 16:9 | `1792x1008`, `2048x1152`, `2560x1440`, `3840x2160` | | 9:16 | `1008x1792`, `1152x2048`, `1440x2560`, `2160x3840` | | 4:3 | `2048x1536`, `4096x3072` | | 3:4 | `1536x2048`, `3072x4096` | | 3:2 | `3072x2048`, `3840x2560` | | 2:3 | `2048x3072`, `2560x3840` | | 3:1 | `1536x512`, `3072x1024`, `3840x1280` | | 1:3 | `512x1536`, `1024x3072`, `1280x3840` | | 21:9 | `2016x864`, `3780x1620` | | 9:21 | `864x2016`, `1620x3780` | | 5:4 | `2560x2048`, `3840x3072` | | 4:5 | `2048x2560`, `3072x3840` | `2000x1001` and other custom dimensions are also outside the standard set; listing every possible size is impractical. **How to fill the fields:** - `size` uses pixel width and height, such as `2048x1152`, separated by a lowercase `x`; the basic example does not additionally send `aspect_ratio`. - For \#77, prefer `1024x1024`, `1536x1024`, `1024x1536`, `1536x864` or the corresponding portrait sizes. - For \#127, prefer 16:9 / 9:16 requests, such as `3840x2160`, `2160x3840`; actual returned images are the source of truth. - Local proxy limits: positive integer dimensions; each side ≤ **16384** and total pixels ≤ **67,108,864**. Multiples of 16 are not required. These limits do not mean every native upstream accepts the same sizes. **Verified two-reference requests:** | Group / model | Requested size | Actual result | |----|----|----| | \#75 / `gpt-image-2` | `2048x1152` | `1672x941`; this request did not use the upscaling account | | \#77 / `gpt-image-2.5-sunburst` | `1536x864` | `1536x864` | | \#127 / `gpt-image-2.5-sunburst` | `3840x2160` | `3584x2016` | All three returned HTTP 200. **Requested dimensions, actual pixels and billing tiers are different concepts. Inspect image dimensions before delivery.**

### Choose your group key, model and size first The commands below are for Bash / WSL. Choose the group you need and execute only the required request. | Group | Example model`MODEL` | Example size`SIZE` | `QUALITY` | |-------|--------------------------|--------------------|-------------| | \#75 | `gpt-image-2` | `2048x1152` | Leave blank | | \#77 | `gpt-image-2.5-sunburst` | `1536x864` | Leave blank | | \#127 | `gpt-image-2` | `3840x2160` | `medium` | Other models in the same group use the same structure; replace `MODEL`. For \#127 flare / sunburst, you may initially omit `quality`; that field does not guarantee a quality-tier switch.

### Text to image: JSON The example below uses \#75. When switching groups, replace the key, `model` and `size`; for \#127's `gpt-image-2` also add `"quality":"medium"`. ``` bash export YM2API_KEY="YOUR_GROUP_75_API_KEY" curl --max-time 300 "https://yyds.seqnode.com/v1/images/generations" \ -H "Authorization: Bearer $YM2API_KEY" \ -H "Content-Type: application/json" \ --data-binary '{ "model": "gpt-image-2", "prompt": "白色陶瓷杯放在木桌上，柔和自然光，无文字", "size": "2048x1152", "n": 1, "response_format": "b64_json" }' \ -o response.json ```

### Image editing: upload two local images The following uses \#127's `gpt-image-2`. For one reference, remove the second `image` field; for other groups, replace parameters according to the table. ``` bash export YM2API_KEY="YOUR_GROUP_127_API_KEY" curl --max-time 300 "https://yyds.seqnode.com/v1/images/edits" \ -H "Authorization: Bearer $YM2API_KEY" \ -F "model=gpt-image-2" \ -F "prompt=保留图1的主体，采用图2的配色，白色背景，无文字" \ -F "image=@reference-1.png" \ -F "image=@reference-2.png" \ -F "size=3840x2160" \ -F "quality=medium" \ -F "n=1" \ -F "response_format=b64_json" \ -o response.json ``` Let cURL generate the multipart headers; do not manually set `Content-Type`.

### Image editing: image URLs / Base64 For this \#77 example, save the following as `edit.json` and substitute actual image URLs or Base64. ``` json { "model": "gpt-image-2.5-sunburst", "prompt": "保留图1的主体，采用图2的配色，白色背景，无文字", "images": [ {"image_url": "https://YOUR_IMAGE_HOST/reference-1.png"}, {"image_url": "data:image/png;base64,YOUR_REFERENCE_2_BASE64"} ], "size": "1536x864", "n": 1, "response_format": "b64_json" } ``` ``` bash export YM2API_KEY="YOUR_GROUP_77_API_KEY" curl --max-time 300 "https://yyds.seqnode.com/v1/images/edits" \ -H "Authorization: Bearer $YM2API_KEY" \ -H "Content-Type: application/json" \ --data-binary @edit.json \ -o response.json ```

### Responses ``` json { "created": 1789380000, "data": [{"b64_json": "BASE64_IMAGE_BYTES"}] } ``` Decode `data[].b64_json` and save the images. When using `response_format=url`, support and link expiration depend on the upstream; download promptly. To retrieve the current key's complete model list, call `GET https://yyds.seqnode.com/v1/models` with the same authentication header.