Mini Program

WeChat Mini Programs call image generation models directly through wx.cloud.extend.AI.createImageModel, without a Cloud Function relay.

Prerequisites

WeChat base library 3.17.4 or later
A CloudBase environment
Image generation resources: the quota included in the Mini Program Growth Plan

Initialization

Initialize CloudBase in app.js. wx.cloud.extend.AI depends on the environment information set during initialization:

App({
  onLaunch() {
    wx.cloud.init({
      env: "your-env-id",
    });
  },
});

Text-to-Image

const imageModel = wx.cloud.extend.AI.createImageModel("hunyuan-image");

const res = await imageModel.generateImage({
  model: "HY-Image-3.0-Plus-4090-Tob-v1.0",
  prompt: "A chubby orange cat napping on a windowsill, watercolor style, warm tones",
  size: "1024x1024",
  revise: { value: true },
  timeout: 120000,
});

console.log(res.data[0].url); // Image URL, valid for 24 hours
console.log(res.data[0].revised_prompt); // Rewritten prompt

Image-to-Image

HY-Image-v3.0-I2I-ToB-v1.0.1 requires one reference image, passed either as image_urls (URL list) or images (base64 list). Use only one of them. A maximum of 1 image is supported, up to 10 MB each, in jpg, jpeg, or png format.

const imageModel = wx.cloud.extend.AI.createImageModel("hunyuan-image");

const res = await imageModel.generateImage({
  model: "HY-Image-v3.0-I2I-ToB-v1.0.1",
  prompt: "Turn this photo into an oil painting",
  size: "1024x1024",
  revise: { value: true },
  image_urls: ["https://example.com/reference.jpg"],
  timeout: 120000,
});

console.log(res.data[0].url);

When the user picks an image from the album, upload it to Cloud Storage first, then use its temporary URL as the reference image:

Page({
  async pickAndGenerate() {
    const { tempFiles } = await wx.chooseMedia({
      count: 1,
      mediaType: ["image"],
    });

    const { fileID } = await wx.cloud.uploadFile({
      cloudPath: `ai-input/${Date.now()}.jpg`,
      filePath: tempFiles[0].tempFilePath,
    });

    const { fileList } = await wx.cloud.getTempFileURL({ fileList: [fileID] });

    const imageModel = wx.cloud.extend.AI.createImageModel("hunyuan-image");
    const res = await imageModel.generateImage({
      model: "HY-Image-v3.0-I2I-ToB-v1.0.1",
      prompt: "Turn this photo into an oil painting",
      image_urls: [fileList[0].tempFileURL],
      revise: { value: true },
      timeout: 120000,
    });

    this.setData({ imageUrl: res.data[0].url });
  },
});

Parameters

AI.createImageModel

function createImageModel(provider: string): ImageModel;

Parameter	Type	Required	Description
`provider`	`string`	Yes	Model provider name, always `"hunyuan-image"`

ImageModel.generateImage

function generateImage(data: GenerateImageInput): Promise<GenerateImageOutput>;

interface GenerateImageInput {
  model: string;
  prompt: string;
  size?: string;
  revise?: { value: boolean };
  footnote?: string;
  image_urls?: string[];
  images?: string[];
  timeout?: number;
}

interface GenerateImageOutput {
  id: string;
  created: number;
  data: Array<{
    url: string;
    revised_prompt?: string;
  }>;
}

Parameter	Type	Required	Description
`model`	`string`	Yes	`HY-Image-3.0-Plus-4090-Tob-v1.0` for text-to-image, `HY-Image-v3.0-I2I-ToB-v1.0.1` for image-to-image
`prompt`	`string`	Yes	Image description, max 500 characters
`size`	`string`	No	Defaults to `1024x1024`. Options: `1024x1024`, `1280x720`, `720x1280`, `1280x1280`
`revise`	`{ value: boolean }`	No	Whether to rewrite the prompt. When enabled, the model optimizes the prompt automatically
`footnote`	`string`	No	Watermark text
`image_urls`	`string[]`	No	Reference image URL list, alternative to `images`
`images`	`string[]`	No	Reference image base64 list, alternative to `image_urls`
`timeout`	`number`	No	Request timeout in milliseconds. `120000` is recommended

The returned data is a list of images. Each item contains url (valid for 24 hours) and revised_prompt (returned when revise is enabled).

Full Example

A working generation page with loading state and error handling:

Page({
  data: {
    prompt: "",
    imageUrl: "",
    revisedPrompt: "",
    loading: false,
  },

  onInput(e) {
    this.setData({ prompt: e.detail.value });
  },

  async generate() {
    const prompt = this.data.prompt.trim();
    if (!prompt || this.data.loading) return;
    if (prompt.length > 500) {
      wx.showToast({ title: "Prompt must be 500 characters or fewer", icon: "none" });
      return;
    }

    this.setData({ loading: true, imageUrl: "", revisedPrompt: "" });

    try {
      const imageModel = wx.cloud.extend.AI.createImageModel("hunyuan-image");
      const res = await imageModel.generateImage({
        model: "HY-Image-3.0-Plus-4090-Tob-v1.0",
        prompt,
        size: "1024x1024",
        revise: { value: true },
        timeout: 120000,
      });

      this.setData({
        imageUrl: res.data[0].url,
        revisedPrompt: res.data[0].revised_prompt || "",
      });
    } catch (err) {
      console.error("Image generation failed", err);
      wx.showToast({ title: "Generation failed, please retry", icon: "error" });
    } finally {
      this.setData({ loading: false });
    }
  },
});

<view class="page">
  <textarea bindinput="onInput" placeholder="Describe the image you want" maxlength="500" />
  <button bindtap="generate" loading="{{loading}}" disabled="{{loading}}">
    {{loading ? "Generating…" : "Generate"}}
  </button>
  <image wx:if="{{imageUrl}}" src="{{imageUrl}}" mode="widthFix" />
  <text wx:if="{{revisedPrompt}}">Rewritten prompt: {{revisedPrompt}}</text>
</view>

Save Images to Cloud Storage

Generated image URLs expire after 24 hours. To keep an image, download it to a local temporary file and upload it to Cloud Storage:

async function saveToCloudStorage(url) {
  const { tempFilePath } = await wx.downloadFile({ url });

  const { fileID } = await wx.cloud.uploadFile({
    cloudPath: `ai-images/${Date.now()}.jpg`,
    filePath: tempFilePath,
  });

  return fileID;
}

⚠️ wx.downloadFile is subject to the downloadFile domain allowlist. Add the image domain in the Mini Program console before release. Displaying a network image with the <image> component is not subject to this restriction.

Compared with Cloud Function Calls

Aspect	Mini Program client	Cloud Function (wx-server-sdk)
Entry point	`wx.cloud.extend.AI.createImageModel`	`cloud.ai().createImageModel`
Base library / dependency	Base library ≥ 3.17.4, nothing to install	wx-server-sdk ≥ 4.0.2
Timeout control	`timeout` parameter (milliseconds)	Cloud Function timeout
API key exposure	None, authenticated by login session	None, server-side call
Best for	Direct generation inside the Mini Program	Persisting results, batch processing, cross-client reuse

Hunyuan Text-to-Image Overview — Model parameters and sizes
wx-server-sdk — Generate images in a Cloud Function
Mini Program Access to AI Models — Text generation with wx.cloud.extend.AI
Hunyuan Text-to-Image 3.0 Upgrade Guide

Prerequisites​

Initialization​

Text-to-Image​

Image-to-Image​

Parameters​

AI.createImageModel​

ImageModel.generateImage​

Full Example​

Save Images to Cloud Storage​

Compared with Cloud Function Calls​

Related​