# Video

URL: /systems/marketing/video · Id: `capability:video`

> Video renders a Video’s timeline into one cut per format, with captions, a transcript and a receipt for every file.

Video renders a [Video](/business/nouns/video)'s timeline into one cut per format, with captions, a transcript and a receipt for every file. Every claim on screen reads from a record.

## Contract
A Video is data, not an edit. Its timeline of scenes and spoken lines is the only timing truth, and picture, voice, music, captions and review all read it. Each scene's picture is captured (a dated [Page](/business/nouns/page) capture or a saved response), drawn from records, or generated.

| Rule | Detail |
| --- | --- |
| Truth ends every scene | Generated picture may open a scene, never close it, and never carries a claim, figure or readable text |
| Generated ceiling | A [Setting](/business/nouns/setting), 20 percent of runtime by default; a [Startup](/business/nouns/startup) can only tighten it |
| Draft mark | Every frame carries it until [finalize](/business/nouns/video/finalize); a Video in `Draft` cannot publish |
| [Voice](/business/nouns/voice) | A spoken line is kept only when its transcript matches the script word for word |
| Sound | Mixed to −14 LUFS and −1.5 dBTP; music never rises inside a marked silence |
| [Review](/business/nouns/review) | Model reviewers propose findings; code checks of transcript, frames, loudness and black frames decide |
| Receipts | Each cut records the timeline [Version](/business/nouns/version), every asset hash and the renderer version |
| Keep everything | Every take, capture and cut is a [File](/business/nouns/file), kept forever |
| Metering | Render seconds and model calls, so a Video's cost is a sum |

| Venue | Cut | Captions |
| --- | --- | --- |
| YouTube | 16:9, 1920×1080, or vertical 9:16 | Sidecar; burned in when vertical |
| Instagram and TikTok | 9:16, 1080×1920 | Burned in |
| LinkedIn | 16:9, 1:1 or 4:5 | Sidecar or burned in |
| X | 16:9 or 1:1 | Burned in |
| The Startup's [Site](/business/nouns/site) | 16:9 embed | Sidecar, with the transcript beside it |

Viewers get MP4 per cut, [Agents](/business/nouns/agent) get a Markdown transcript with scene list and sources, and clients get the timeline as JSON. The [Media production](/systems/media) [System](/business/nouns/system) publishes and measures what this capability renders.

## Verbs
| Verb | Object | Event |
| --- | --- | --- |
| [Video.draft](/business/nouns/video/draft) | [Video](/business/nouns/video) | [`Video.drafted`](/business/events/Video.drafted) |
| [Video.finalize](/business/nouns/video/finalize) | [Video](/business/nouns/video) | [`Video.finalized`](/business/events/Video.finalized) |

## Interfaces
- [API](/code/api/reference)
- [SDK](/code/sdk/reference)
- [CLI](/code/cli/reference)
- [MCP](/code/mcp/reference)

## Example
`slotwise.app` sells team scheduling. An Agent drafts a 90-second explainer from its [Offer](/business/nouns/offer) and [Brand](/business/nouns/brand), capturing the pricing Page and a live API response. It renders a 16:9 cut for YouTube and a 30-second 9:16 cut for LinkedIn and TikTok.