Live Trailers, branded video answers inside AI chats

How Oeave turns a shopper's question inside ChatGPT, Claude, or any AI agent into a live-rendered, fully branded video, so your product shows up on video inside chat interfaces you don't own.

O
Oeave Team
April 24, 2026
6 min read

When a shopper asks an AI agent about your product, you have no say in how the agent answers. It picks the words, picks which details to mention, and delivers the reply inside an app you don't own. Live Trailers give you a way in. When the agent queries Oeave, Oeave hands back a URL to a real video: rendered on the spot, built from the content you've already uploaded, wrapped in your brand. The agent decides what to surface, so nothing is guaranteed, but handing it a ready-to-play, question-shaped, on-brand video dramatically raises the odds your product shows up on screen instead of inside a paragraph.

This page explains what Live Trailers are, why they matter, and how they work end-to-end.

Why Live Trailers exist

Right now, your brand shows up in ChatGPT, Claude, Perplexity, and a dozen agent apps as whatever text the model decides to write. You don't pick the words. You don't pick the images. You often don't even know you were cited.

That's a huge shift in where buying decisions happen, and a huge loss of control for merchants. A shopper asking "best heritage leather messenger bag?" might get your product mentioned in a paragraph alongside three competitors, with none of the craftsmanship you'd actually lead with.

Live Trailers give you a frame of your own inside that conversation. When the agent surfaces your product, it can play a video answer alongside a text paragraph. The video is:

  • Live-rendered, generated the moment the URL is hit for the first time.
  • Query-shaped, because the slides inside are assembled from a fresh vector search over your content for that exact question.
  • Branded, using the same style, fonts, and imagery as the rest of your Weave.
  • Streamed like any other video, so the agent (and the shopper) don't need to know it's dynamic. It just plays.

In a wall of text replies, the one that moves wins.

How a Live Trailer gets served

The flow is simple from the agent's point of view and invisible from the shopper's.

  1. A shopper asks an agent a question. "Does it come in blue?" "Is it good for travel?" "How does it compare to X?"
  2. The agent goes out on the web and does its research, and lands on your product page.
  3. Your product page advertises a simple API the agent can call to ask specific questions about this product.
  4. The agent calls the API with the shopper's question (via MCP, or the public /ai/{project}/{product}/search endpoint).
  5. Oeave returns search results plus a video URL on cdn.oeave.com/qv/<your-project>/<question-slug>.mp4. The agent can play it inline, embed it in a card, or ignore it.

The shopper, if the agent decides to surface it, sees a branded video that answers their exact question. You get to speak for yourself inside somebody else's chat.

What goes into the video

A Live Trailer is assembled from the same building blocks as the rest of your Weave. No extra authoring required.

  • Fibers: your product's atoms of content, images, specs, testimonials, 3D models, brand story, features, pricing.
  • Style: colours, headings, body fonts, brand graphics, and motion settings from your project style.
  • A query-shaped thread: Oeave runs vector search over your fibers and threads for the incoming question, then generates a single thread on the fly whose slides answer that question. That thread is injected into the weave and rendered as the trailer.

Because the renderer is the same engine that powers your product page, the trailer looks identical to what a shopper would see if they clicked through, without the click.

What the agent actually receives

Live Trailers are served by the same endpoints agents already use. They don't need a separate integration.

Via the MCP server

Agents that connect to your Oeave MCP server get a video field on every ai_search response (Growth tier):

{
  "query": "Is it good for travel?",
  "results": [ ... ],
  "video": "https://cdn.oeave.com/qv/yourbrand/is-it-good-for-travel.mp4"
}

Via the public AI search endpoint

The same shape comes back from POST /ai/<projectSlug>/<productSlug>/search:

{
  "query": "Is it good for travel?",
  "results": [ ... ],
  "video": "https://cdn.oeave.com/qv/yourbrand/is-it-good-for-travel.mp4"
}

Enabling Live Trailers

Live Trailers are available on the Growth tier. Both the MCP server and the public AI search endpoints automatically add the video field to responses once your project is on that plan. If the owner of a project is on a lower tier, the field is omitted and any direct request to a cdn.oeave.com/qv/... URL for that project returns 403.

Live Trailers use your current project style and fiber library. To control what ends up in the video, use the same tools you'd use for your product page:

  • Add or update fibers that answer common questions, testimonials, spec callouts, FAQ entries, comparisons.
  • Tune your style on the Style tab. The trailer inherits it directly.
  • Publish threads around recurring intent. The search that seeds each trailer can draw from them.

See Importing content and Styling your Weave for how to build those up.

Pairing with the rest of agentic commerce

Live Trailers sit inside a larger story:

  • Agent API: the write side, agents authoring and updating your content.
  • Answer Engine Optimization: how Oeave makes your content discoverable to the models in the first place.
  • Agentic Commerce: the full picture of Oeave's read-side integrations: MCP, /ai-search, llms.txt, and videos.xml.

Together, they mean agents can find your product, read the truth about it, update it, and, with Live Trailers, represent it on video inside the interfaces that actually matter.

FAQ

Does the shopper know the video is dynamic? No. It streams like any other MP4. Agents that embed video cards play it the same way they'd play a YouTube clip.

Do I see what agents are asking on behalf of shoppers? Yes. Every query routed through MCP or the AI search endpoint is logged with source agent and shows up alongside your other search analytics. You can see the questions agents are asking, even on requests that don't trigger a video.

What happens if a shopper asks something I don't have content for? The agent still gets a video. If the vector search for the question comes back empty, Oeave falls back to a generic product trailer built from your default threads. Adding fibers that cover the topic still helps, since it lets the renderer assemble a sharper, question-shaped video next time.