Decision from images - HeyGen Documentation

Add up to 4 images under assets.images to decide on what something looks like: a product photo against a spec, a thumbnail against brand rules, or a screenshot against an expected state. Images are keyed by names you choose. The model reads all images in request order, then state, so use state to say what each image is by its position.

Compare two images

curl -X POST https://api.heygen.com/v3/models/decisions \
  -H "X-Api-Key: $HEYGEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "heygen-decision-1",
    "state": "The first image is the approved thumbnail template. The second is a new thumbnail submitted for review.",
    "assets": {
      "images": {
        "template": { "type": "asset_id", "asset_id": "TEMPLATE_IMAGE_ASSET_ID" },
        "submission": { "type": "url", "url": "https://example.com/thumbnail.png" }
      }
    },
    "questions": {
      "on_brand": {
        "type": "noul",
        "instructions": "Does the second image follow the layout, colors, and logo placement of the first?"
      },
      "text_legibility": {
        "type": "score",
        "instructions": "How readable is the title text in the second image at small sizes?",
        "criteria": [
          "Title text is missing or unreadable",
          "Title text is readable only at full size",
          "Title text is clear at thumbnail size"
        ]
      }
    }
  }'

The response has the same data.answers and data.usage fields as a text request.

Image inputs


Field	`assets.images`, keyed by name
Per request	Up to 4 images
Sources	`asset_id`, HTTPS `url`, or `base64` with an `image/*` `media_type`. See Choose an asset source.
Size	32 MiB per URL download, 5 MiB per decoded base64 image

Images can share a request with one video. Each image adds to usage.input_tokens.

⌘I