Decision from video - HeyGen Documentation

Add one video under assets.videos to decide on what appears on screen: whether a product is shown correctly, whether a scene matches a reference, or how clearly something is visible. The model reads the video as sampled frames, after any images and before state. To judge what is said, put the transcript in state.

Check a video against a reference image

This request compares the package in a product demo with the approved design.

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 image shows the approved package design. The video is a product demo. Compare the package visible in the video with the approved design.",
    "assets": {
      "images": {
        "approved_package": { "type": "asset_id", "asset_id": "IMAGE_ASSET_ID" }
      },
      "videos": {
        "demo": { "type": "asset_id", "asset_id": "VIDEO_ASSET_ID" }
      }
    },
    "questions": {
      "package_matches": {
        "type": "noul",
        "instructions": "Does the package shown in the video match the approved design in the image?"
      },
      "logo_visibility": {
        "type": "score",
        "instructions": "How clearly is the package logo visible in the video?",
        "criteria": [
          "The logo is not visible",
          "The logo is visible but too small or obscured to read",
          "The logo is clear and readable"
        ]
      }
    }
  }'

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

Video inputs


Field	`assets.videos`, keyed by name
Per request	1 video, with up to 4 images
Sources	`asset_id`, HTTPS `url`, or `base64` with a `video/*` `media_type`. See Choose an asset source.
Size	50 MiB per URL download, 16 MiB per decoded base64 video, within the 16 MiB request body

For larger files, upload with the asset upload API and pass the asset_id. Sampled frames add to usage.input_tokens.

⌘I