Overview

Pair a pre-recorded video with a live D-ID agent using a single script tag

What is Agentic Videos

Agentic Videos pairs a pre-recorded video with a live D-ID agent. A viewer watches the video and clicks Ask at any time to start a real-time conversation with an agent grounded in that video's content. When the conversation ends, they return to the video where they left off.

Add it to any page with a single <script> tag. No backend code, no bundlers, no build steps.

When to use

  • You have existing video content and want to make it interactive.
  • You want viewers to ask questions about what they're watching.
  • You want the quickest integration (minutes) with a pre-built player.
  • You want the agent rendered inline, in your own page layout.

Key concepts

Video-first

The experience starts as a normal video player — native video or YouTube. Playback controls are built in.

Ask to converse

An Ask pill appears while the video plays. Clicking it crossfades to a live agent the viewer can talk to by voice or chat, at any point in the video.

Grounded agent

The agent is grounded in the video's script and supplemental knowledge, so answers stay on-topic for what the viewer is watching.

Back to video

When the viewer finishes the conversation, the player crossfades from the agent back to the video and resumes playback.


How it works

The embed script self-discovers its own tag, reads its data-* attributes, fetches the agentic-video resource, and renders the player. Once the video has finished processing, the D-ID agent runtime is injected in the background during idle time so it's ready by the time the viewer clicks Ask.

---
config:
  theme: redux
  look: neo
---
flowchart LR
    Page(["fa:fa-globe Your Page"]) --> Script(["fa:fa-code Embed Script"])
    Script --> Video(["fa:fa-play Video Player"])
    Video -- "fa:fa-comment Ask" --> Agent(["fa:fa-robot D-ID Agent"])
    Agent -- "fa:fa-arrow-left Back" --> Video

The agent flow activates once the agentic-video has finished processing (status: done). While it's still processing — the video plays on its own, with no Ask pill.

How the agent is built

Creating an agentic video runs the source video through a pipeline. D-ID transcribes the audio, summarizes the transcript into the agent's knowledge base, and splits it into a timeline of topics — each tagged with the second it begins. The agent is grounded in that knowledge, so it answers from what the video covers, and it uses the timeline to jump the viewer to the right moment when asked.

The result is a standard D-ID agent. You can refine it at any time — instructions, knowledge, voice, and tools — by editing the agent directly.

Use cases

  1. Learning and training: Turn a recorded lesson into a self-paced session where learners ask for clarification as they watch.
  2. Product marketing: Let viewers explore a feature walkthrough and ask follow-up questions inside the player.
  3. Pre-sales: Pair a demo video with an agent that answers prospect questions on demand.
  4. Customer support: Pair a how-to video with an agent that guides the user through the specific questions the video raises.

Get started

Ready to add an agentic video to your site? Follow the Quickstart guide to integrate in a few minutes.

FAQ