Hunyuan Large Model User Guide

Image-to-Text Generation

tip

Call createModel and pass "hunyuan-exp" or "hunyuan-open" to create the Hunyuan Large Model, and use "hunyuan-vision" as the model parameter passed in when calling streamText / generateText.

For details, please refer to the sample code below.

When using Image-to-Text Generation, the parameters passed differ from those in ordinary dialogue text generation. messages[n].content is an array, and the elements in the array can be objects representing text/images, as shown below:

async function readImage() {
  const model = ai.createModel("hunyuan-exp");
  const res = await model.streamText({
    model: "hunyuan-vision",
    messages: [
      {
        role: "user",
        content: [
          {
            type: "text",
            text: "What is the content of the image below?",
          },
          {
            type: "image_url",
            image_url: {
              url: "https://cloudcache.tencent-cloud.com/qcloud/ui/portal-set/build/About/images/bg-product-series_87d.png",
            },
          }, 
        ],
      },
    ],
  });

  for await (let x of res.textStream) {
    console.log(x);
  }
}

When messages[n].content represents text, the object type is expressed as:

interface TextContent {
  type: "text", // fixed value
  text: string, // Text content
}

When messages[n].content represents an image, the object type is expressed as:

interface ImageContent {
  type: "image_url", // fixed value
  image_url: {
    url: string, // Image URL
  },
}