Building an AI‑Powered Image Generation Mini‑Program with Go Backend and Tencent Cloud

This article describes how to create a mini‑program that generates images from user‑provided text using AI. The overall workflow consists of three parts: (1) text‑to‑image search, (2) image style‑transfer, and (3) a simple HTTP service that ties the two together.

Text‑to‑image search – The author uses Sogou’s picture search API. By sending a keyword (and optional tag filters) to https://pic.sogou.com/napi/pc/searchList the API returns a list of image URLs. The Go implementation of the request is shown below:

params := url.Values{}
params.Set("mode", "1")
params.Set("start", "0")
params.Set("xml_len", "48")
params.Set("query", keywords)
// optional tag filter handling omitted for brevity
uri := "https://pic.sogou.com/napi/pc/searchList"
address, _ := url.Parse(uri)
address.RawQuery = params.Encode()
request, _ := http.NewRequestWithContext(ctx, http.MethodGet, address.String(), nil)
resp, _ := http.DefaultClient.Do(request)
body, _ := ioutil.ReadAll(resp.Body)
var rsp Response
json.Unmarshal(body, &rsp)

The response structure is defined in Go (see the long type Response struct { … } in the source).

Image style‑transfer – After obtaining a suitable picture, the author calls Tencent Cloud’s FaceCartoonPic API to convert the image into a cartoon style. The required SDK calls are:

credential := common.NewCredential("***", "***")
cpf := profile.NewClientProfile()
client, _ := ft.NewClient(credential, "ap-guangzhou", cpf)
req := ft.NewFaceCartoonPicRequest()
req.Url = common.StringPtr(imageUrl)
req.RspImgType = common.StringPtr("url")
resp, err := client.FaceCartoonPic(req)
if err != nil { /* handle error */ }
return []byte(*resp.Response.ResultUrl), nil

Both the request/response protobuf definitions and the Go wrapper functions ( FaceCartoonPicPro , SearchImage ) are provided.

HTTP service – The backend exposes a single endpoint /SearchImage that accepts a JSON payload containing the text and optional tag JSON. It performs the search, calls the style‑transfer API on the first successful image, and returns a list of original and result URLs.

http.HandleFunc("/SearchImage", func(w http.ResponseWriter, r *http.Request) {
    data, _ := ioutil.ReadAll(r.Body)
    req := &pb.SearchImageReq{}
    json.Unmarshal(data, &req)
    rsp := &pb.SearchImageRsp{}
    SearchImage(context.Background(), req, rsp)
    out, _ := json.Marshal(rsp)
    w.Write(out)
})
http.ListenAndServe("127.0.0.1:8080", nil)

WeChat mini‑program front‑end – The UI consists of a simple input box, a button, and a canvas to display the generated image. Key files are:

WXML (layout):

<view class="container">
  <div class="form-item" style="width:673rpx;height:70rpx;">
    <input placeholder="写下你的创意" bindinput="handlerInput" />
    <button loading="{{buttonStatus}}" bindtap="handlerSearch">立即生成</button>
  </div>
  ...
  <canvas type="2d" id="input_canvas" style="background:#e4e4e1;width:673rpx;height:700rpx;"></canvas>
</view>

JS (logic):

Page({
  data:{ inputValue:"", tags:[], option:[], buttonStatus:false },
  handlerInput(e){ this.setData({inputValue:e.detail.value}) },
  handlerSearch(){ if(this.data.inputValue.length===0){wx.showToast({icon:"error",title:"请输入你的创意"});return;} this.imageDraw(); },
  imageDraw(){
    const opt = this.data.option.length?{tags:this.data.option}:{};
    wx.request({
      url:'http://127.0.0.1:8080/SearchImage',
      method:'POST',
      data:{text:this.data.inputValue, option_json:JSON.stringify(opt)},
      success:res=>{ /* parse response, download result image, draw on canvas */ }
    })
  }
})

The demo shows successful generation of cartoon‑style images for several sample keywords, with screenshots included in the original article.

In conclusion, the author demonstrates a complete end‑to‑end solution for AI‑driven image creation: a searchable image repository, a cloud‑based style‑transfer service, a Go backend that orchestrates the workflow, and a WeChat mini‑program front‑end for user interaction.