Webinars at scale

Webinars are a heavy production. A traditional webinar is a week of work: outline, deck, dry runs, recording, edit, follow-ups. 10ex compresses that production cost so a single operator can run many.

Pipeline

webinar-generator (Manager-level crew) runs six sub-agents in sequence:

1. brand_rag_agent. Pulls brand context from Knowledge.
2. outline_agent. Generates a content outline against an audience segment.
3. script_agent. Writes the spoken script.
4. pptx_agent. Renders slides.
5. avatar_agent. Generates an AI host (HeyGen or similar).
6. stitch_agent. Composites slides plus avatar into a finished video.

Output: a Zoom-style replay-ready video, a registration landing page, an email funnel, and post-webinar follow-up sequences.

When this is the right tool

• You want a webinar this week, not next quarter.
• You’re testing a topic before committing a presenter to a live recording.
• You want a localized variant per region without re-recording.

A common gotcha: the AI host is convincing, but it isn’t a substitute for a live, interactive session. Use the generated webinar as evergreen replay content, then layer live Q&A on top when the topic warrants it.

Common questions

How long does a full run take? Most webinars finish within 30 to 60 minutes end-to-end, dominated by avatar video generation.

Can I edit the script before the avatar records? Yes. Atlas pauses after the script step (configurable) so you can review before the expensive video render.

What languages are supported? English plus the language set HeyGen ships. New languages roll in as the underlying provider expands.

Related

• Use cases, Webinars
• webinar-generator reference
• Approval workflows: how Atlas gates the calendar invite step