media

Overview

The media parent type covers audio, video, and multimedia content including videos, podcasts, music, image galleries, livestreams, and animations. It captures the metadata agents need for media discovery, indexing, and accessibility.

Subtypes

Subtype	Qualified Type	Description
`video`	`media.video`	Video content (YouTube, Vimeo, etc.)
`podcast`	`media.podcast`	Podcast episode or series page
`music`	`media.music`	Music track, album, or playlist
`image_gallery`	`media.image_gallery`	Photo or image collection
`livestream`	`media.livestream`	Live streaming content
`animation`	`media.animation`	Animated or interactive media

type_data fields

Field	Type	Required	Description
`title`	`string`	Yes	Media title
`format`	`string`	No	Media format (video, audio, image, interactive)
`duration`	`string`	No	Duration in ISO 8601 duration format (e.g., “PT1H23M45S”)
`creator`	`string`	No	Creator or channel name
`platform`	`string`	No	Hosting platform (youtube, spotify, etc.)
`transcript_available`	`boolean`	No	Whether a transcript is available

Example type_data

{
"type_data": {
  "title": "Attention Is All You Need — Paper Explained",
  "format": "video",
  "duration": "PT45M12S",
  "creator": "Yannic Kilcher",
  "platform": "youtube",
  "transcript_available": true
}
}

Notes

title is the media-specific title, which may differ from metadata.title (the page title).
format normalizes to: "video", "audio", "image", "interactive". This is the primary media format, not the file format.
duration uses ISO 8601 duration format. "PT1H23M45S" means 1 hour, 23 minutes, 45 seconds. For content without a fixed duration (e.g., livestreams), omit this field.
creator is the content creator or channel name, not the platform. Use platform for the hosting service.
platform is normalized to lowercase identifiers: "youtube", "spotify", "vimeo", "soundcloud", "twitch", etc.
transcript_available indicates whether a transcript exists, not whether the SDF document includes it. Transcripts, when available, are in the sections field.
For podcast episodes that are essentially interviews, use aspects: ["discussion.comment_thread"] if the conversational structure is significant.