Skip to content

Hy HOT โ€‹

Updated

Endpoint: POST https://vod.intl.tencentcloudapi.comAction: CreateAigcImageTask

Hy Image is Tencent Hunyuan's image generation model, supporting text-to-image and reference-image generation / editing. 3.5-preview supports multi-turn re-editing based on the previous result.

Basic information โ€‹

ItemValue
ModelNameHunyuan
ModelVersion3.5-preview / 3.0
DefaultModelVersion=3.5-preview
Supported modesText generation, reference image generation, re-editing of generated output
Guardrail bypass supportNot supported

The input and output guardrail bypass parameters do not apply to Hy Image.

It supports both text generation and reference-image-based generation.

3.0 accepts 3 reference images, and 3.5-preview accepts up to 6.

Request parameters โ€‹

ParameterRequiredTypeDescription
ModelNameRequiredStringFixed value Hunyuan
ModelVersionOptionalString3.5-preview / 3.0
PromptRequiredStringThe generation prompt. When a reference image is present, it acts as the editing instruction
FileInfos.NOptionalArrayReference images. Put Type and Url in it
OutputConfig.ResolutionOptionalString1K / 2K / 4K
OutputConfig.AspectRatioOptionalString1:1 / 16:9 / 9:16 / 4:3, and so on
OutputConfig.PersonGenerationOptionalStringAllowAdult / Disallowed
ExtInfoOptionalStringSpecifies the pixel size directly. See the entry below

Output images per request

3.5-preview generates 1 image per request. For multiple images, submit requests sequentially.

Request examples โ€‹

This is text generation.

json
{
  "SubAppId": 123456789,
  "ModelName": "Hunyuan",
  "ModelVersion": "3.5-preview",
  "Prompt": "a serene japanese garden, soft morning light",
  "OutputConfig": {
    "Resolution": "2K",
    "AspectRatio": "16:9",
    "StorageMode": "Temporary"
  },
  "SessionContext": "job-001"
}

This is reference-image-based generation. When you put in a reference image, Prompt is interpreted as an editing instruction.

json
{
  "SubAppId": 123456789,
  "ModelName": "Hunyuan",
  "ModelVersion": "3.5-preview",
  "Prompt": "change the car color to matte black, keep the composition",
  "FileInfos": [
    {
      "Type": "Url",
      "Url": "https://<host>/car.png"
    }
  ],
  "OutputConfig": {
    "Resolution": "1K",
    "AspectRatio": "16:9",
    "StorageMode": "Temporary"
  },
  "SessionContext": "job-001"
}

Response example โ€‹

json
{
  "AigcImageTask": {
    "Status": "FINISH",
    "ErrCode": 0,
    "Progress": 100,
    "Output": {
      "FileInfos": [
        {
          "FileUrl": "https://store.vod-qcloud.com/.../aigcImageGenFile.png",
          "ExpireTime": "2026-10-02T07:59:41Z",
          "MetaData": {
            "Width": 2048,
            "Height": 1152,
            "Container": "png_pipe"
          }
        }
      ]
    }
  }
}

Special settings โ€‹

Output resolution โ€‹

These are the actual output pixels for each combination of Resolution and AspectRatio. Values are aligned to multiples of 16.

Resolution1:116:99:164:33:4
Omitted720x7201280x720720x1280960x720720x960
1K1024x10241808x10241024x18081360x10241024x1360
2K2048x20482048x11521152x20481920x14401440x1920
4K4096x40963840x21602160x38403840x28802880x3840

If you omit Resolution and pass only AspectRatio, the short side is fixed to 720. If you omit both values, the output is 1872x1248 at a 3:2 ratio. Even with the same Resolution, the actual pixels vary by ratio, so if you need an exact size, specify it directly with ExtInfo below.

The 4K / 5:4 combination is output as 2688x2160.

Specifying the pixel size directly โ€‹

When you put size in ExtInfo, it takes precedence over Resolution and AspectRatio. The value is in widthxheight format.

json
{
  "ExtInfo": "{\"AdditionalParameters\": \"{\\\"size\\\":\\\"1808x1024\\\"}\"}"
}
Input valueActual output
1808x10241808x1024
656x1024656x1024
656x1000656x992 (rounded down to a multiple of 16)
autoDetermined automatically from the prompt
100x100256x256 (corrected to the lower bound)

The output size is corrected against multiples of 16 and the minimum size. When auto is specified together with Resolution, the ratio is determined based on that resolution; on its own, the ratio is determined from the prompt.

Reference images โ€‹

Specify Type and Url in FileInfos. For image files, set Type to Url.

json
{
  "FileInfos": [
    {
      "Type": "Url",
      "Url": "https://<host>/reference.png"
    }
  ]
}
ItemValue
Allowed fieldsType(Url), Url / ReferenceType (optional)
Number of referencesUp to 6
Formatsjpeg, png, webp

Reference image fields

FileInfos specifies Type="Url" and the image Url.

Operational notes โ€‹

ItemValue
1K processing time21 seconds at 12 seconds
2K processing time30 seconds at 21 seconds
4K processing time46 seconds at 34 seconds

Multi-image requests are processed sequentially. Confirm Status="FINISH" / ErrCode=0 and Output.FileInfos in DescribeTaskDetail before submitting the next request.

Built with VitePress ยท Deployed on Tencent EdgeOne Pages