How do you make an AI explainer video?

Write a 60–120 second script. Generate the voiceover with ElevenLabs. Generate matching b-roll on AI video (one clip per script paragraph). Drop everything into CapCut or Descript, add on-screen text for key terms, export. Total production time for a 90-second explainer is 30–60 minutes once you have a template.

Script → voiceover → b-roll → edit → publish.
One b-roll clip per paragraph of script.
On-screen text for jargon and numbers.
Keep total runtime under 90s for best retention.

The script structure that converts

Hook (0–5s): the problem in one sentence. Body (5–60s): the explanation in 2–4 simple steps. CTA (60–90s): what to do next. Each beat = one b-roll clip. Write the script before touching any AI tool — saves hours of misaligned generation.

Picking b-roll that actually helps

Literal illustrations beat abstract aesthetics. If the script says 'open the app', show a hand opening a phone. Avoid 'pretty AI loops' that have nothing to do with the words being spoken — retention drops sharply.

FAQ

Can I make explainer videos without a voiceover?

Yes, but engagement is lower. Captioned silent videos are best for muted feeds (TikTok, LinkedIn); voiceover is best for YouTube.