Multimodal inputs

Multimodal input allows a supported model to process text, images, and other content together. This page uses a text-and-image chat request. It does not imply that every model supports images, audio, or video.

Choose a model

Before sending a request, confirm that:

• The model supports image input.
• The model supports the Chat Completions interface.
• Image format, count, size, and resolution meet its limits.
• The current API key can access the model.

Text and image example

{
  "model": "YOUR_MULTIMODAL_MODEL_ID",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Describe the main content of this image."
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "https://example.com/image.jpg"
          }
        }
      ]
    }
  ]
}

This is an example compatible structure. Support for URLs, Base64 data, or other input methods depends on the selected model.

Prepare an image

A local file path or a webpage that requires a browser login usually cannot be used as a remote image input. Do not make a private file public only to call the API.

Control usage

Images may use a different metering method from ordinary text. Remove unnecessary images and repeated history, then review media usage and cost according to the model documentation.

Troubleshooting

• The image cannot be read: confirm that the URL returns an image and is reachable from the service.
• The format is unsupported: convert the file to a supported format instead of changing only its extension.
• The request is too large: compress, resize, or remove images.
• The response structure differs: follow the interface documentation for the selected model.