How to make a video ad with AI: brief, draft, approve, reuse

To make a video ad with AI, write a brief that names the deliverable (length, vertical or horizontal, presenter or not, captions) and the product, give it to a video agent, and ask for a draft before it spends on the final render. On Sume, that's a thread at sume.com/agents: the agent picks the models and asks before it spends, and you can save the finished recipe as a Format so the next ad is one request.

The steps follow Sume's Quick start and Sume basics, read on 2026-09-27. If "video agent" is new to you, What is a video agent? explains the term.

What should a video ad brief say?

Describe the deliverable, not the tool calls: that is how Sume's quick start puts it, because the agent picks the models. A useful brief answers five questions:

• Where will it run? The placement sets the shape (9:16, 1:1, or 16:9) and the length; how long an ad should be lists each platform's numbers.
• What is the product, what is the one message, and what should the viewer do next?
• Presenter or not: a talking presenter, a creator-style clip, or product shots only?
• Captions and music: do you want burned-in captions or a music bed?
• Where is the checkpoint? Ask for a draft before the final render.

Make me a 15-second vertical product promo for my product. Use a presenter
avatar, add captions, and show me a draft before spending credits on the final
render. Once I approve it, save this thread's recipe as a Format called
product-promo so I can call it from my backend.

How do I make the ad in Sume's Agents tab?

The brief above is the one Sume's quick start suggests. From there:

• Open sume.com/agents, sign in, and start a thread. Nothing to install.
• Paste the brief into the composer. The thread keeps every artifact and approval.
• The agent picks the models and asks before it spends. A recipe written for chat can also pause at a quality gate, such as approving stills before it makes the video.
• Review the draft, ask for changes in the same thread, and approve the spend for the final render. The docs describe the chat as the place to write a brief, approve spend, inspect artifacts, and shape a house style over a few turns.

Can a ready-made Format make the ad instead?

Sume also ships catalog Formats, saved recipes you call at the reserved sume handle with POST /v1/formats/sume/{slug}/runs. Four describe ad work. A description is the Format's stated aim, not a promise about any one video; the full list is in Ready-made Formats for product video.

How do I make the next ad without starting over?

Ask the agent to save the thread's recipe as a Format. A Format is that recipe, addressable by handle and slug, and it is what your backend calls for the next product: POST /v1/formats/{handle}/{slug}/runs with a new instruction.

A run over the API is unattended: approvals the chat would ask for are pre-granted, and the run carries on within its spend cap. It can carry up to 30 product images as attachments, each a public HTTPS URL. What is a Sume Format? covers the call, its API key scopes, spend caps, and bulk runs.

What does an AI video ad cost?

There is no fixed price per ad: it depends on the models and the length the agent uses. Generation is metered at the rates on API pricing, plus a 5.5% agent fee by default, and the wallet also pays for the agent's own turns. GET /v1/usage with a thread_id sums exactly what one thread cost.

Before the ad runs, check it against the platform's spec (for example Facebook in-stream), compare the product in stills with your photos (why frames matter), and confirm your rights to any claim or music in it.