Uncover AIGuides
  1. Guides
  2. Tool guides
  3. How to make an explainer video with AI
Guide · AI video

How to make an explainer video with AI

From a one-line idea to a finished video: script, scenes, visuals, voice and edit, with the prompts and tools for each step. Or one click from your documents if good enough is enough.

Founder, Uncover AI
Updated Facts from vendor pricing and help pages, checked 23 Oct 2026

Nobody paid to appear on this page, and we earn nothing from the links on it.

Logos are trademarks of their owners. Initials stand in where no logo is available to us.

The workflow at a glance

An AI explainer video is five jobs: write the script, plan the scenes, generate the visuals, add the voice, and edit. Use a chat assistant for the first two, a video or screen tool for the visuals, an AI voice for narration, and a transcript-based editor to finish. The first video is the slow one; after that you reuse the same style prompt, voice and edit template.

There is also a one-click route. If your explainer summarizes documents you already have, Gemini Notebook (the new name for NotebookLM) turns them into a narrated Video Overview with no editing at all.1 It is the right call for internal briefings and course material. For anything with your brand on it, use the five steps.

  1. Write the scriptA chat assistant drafts it; you cut it to the length and hook you need.
  2. Plan the scenesTurn every line into a shot: visual, on-screen text, duration.
  3. Generate visualsAI clips, screen recordings or motion graphics, one scene at a time.
  4. Add the voiceGenerate narration from the final script with an AI voice.
  5. Edit and exportAssemble, caption and export in the sizes you need.

The fast route: a Video Overview from your sources

Open a notebook with your sources, select Video Overview in the Studio panel, then pick a format, language, visual style and an optional prompt to focus it. Formats are Explainer, Short (about 60 seconds) and Cinematic (English only). Styles include Whiteboard, Watercolor, Paper-craft and Anime, or describe your own. Generation can take more than 30 minutes, and you download the file when it is done.1 Google notes the voices and visuals may contain inaccuracies, so watch it before you share it.

For a research-led explainer, our NotebookLM guide covers how to set up the notebook.

Write the script

The script decides whether the video works; visuals cannot rescue a vague one. Keep it short: around 140 to 160 spoken words is a minute of narration. Open with the problem in one sentence, explain one idea, and end with one action. Any strong chat assistant can draft it. The prompts below work in Claude, whose Pro plan is $17 a month billed yearly, or in ChatGPT and Gemini.2

Write a 60-second explainer video script about [topic] for [audience]. Split it into six 10-second scenes. Open with a one-sentence hook that names the problem the viewer has. One idea per scene, short spoken sentences, no jargon. End with one clear next step. Keep it under 150 words, and add a one-line visual idea under each scene.

Tip: Read the script out loud before you do anything else. Every sentence you stumble on, the AI voice will stumble on too.

Plan the scenes

Turn the script into a shot list before you generate anything. A shot list is cheap to change; a generated clip is not. For each scene, write the line of narration, what the viewer sees, any on-screen text, and the length in seconds.

Then fix a style and keep it for every scene. A method we covered in July 2026: give an assistant a few reference videos you like, ask it for a written breakdown of their editing style, and have it produce one reusable "style prompt" you paste into every generation. That one step is what stops the video looking like six unrelated AI clips.

Here is my 6-scene script. For each scene, write: (1) the narration line, (2) a visual description for an image model in this style: [paste style prompt], (3) on-screen text of five words or fewer, (4) duration in seconds. Keep the same character, palette and camera style in every scene. Do not put text inside the images; I will add it in the edit.

Generate visuals

Pick the kind of visual that fits the subject. Most explainers mix two of these three.

AI-generated scenes

For illustrated or cinematic scenes, generate a still image for each scene first, then animate it into a short clip. Images are cheaper to redo than video, and starting from an approved frame keeps the look consistent. To make scenes flow into each other, use the last frame of one clip as the reference for the next.

Kling's current video model is Kling 4.0.4 Its Standard plan is $8.80 a month after a discounted first month, with 660 credits a month.5 If you want several models in one place, FLORA puts Veo 3.1, Seedance 2.0, Kling and 60+ other models on one canvas; video models start on the $18-a-seat Starter plan, and the free plan covers images only.3 Google AI Pro, $19.99 a month, includes 1,000 Google Flow credits for Google's own video models.6, 7

Screen recordings

For a product or software explainer, a clean screen recording beats any generated clip. Tella records polished screen videos with auto zoom, layouts and filler-word removal, from $13 per user a month.12 Cap is a free, open-source option for personal use; Cap Pro adds AI titles, summaries and chapters for $8.16 a month billed yearly.13

Motion graphics

For charts, counters and animated text, have a coding agent build them. HyperFrames is a free, open-source framework that turns HTML and animation code into MP4 video, built for agents such as Claude Code to write.14 You describe the graphic in plain English; the agent writes and renders it. Our YouTube editing guide walks through that setup.

The agent route: one sentence to a full video

In July 2026 we covered a Claude Code skill that takes one sentence and produces a complete narrated explainer, with Higgsfield generating the visuals. Higgsfield connects to Claude as a custom connector: Settings, Connectors, Add custom connector, with the server URL https://mcp.higgsfield.ai/mcp. It needs a paid Higgsfield subscription, and every generation uses credits.8 In that walkthrough, asking for a Spanish version reused the same visuals with a new voiceover, which makes localizing an explainer nearly free.

Add the voice

Generate the narration from the final script, not a draft, and do it before the edit so the cuts can follow the voice. Check the commercial license: on most free plans you cannot use the audio in anything you sell or advertise.

Rewrite this script for an AI voiceover. Spell out numbers and acronyms the way they should be said, break long sentences in two, and mark one word per sentence to stress. Keep the meaning and the length.

If you want an on-screen presenter instead of a voice over visuals, Hedra's Character 3 model turns an image plus audio into a lip-synced talking host through Hedra's API16, and its Play app has plans from $20 a month with commercial use.15 Our AI voice generators guide compares more voice tools.

Edit and export

Lay the voiceover down first, then cut each scene to it. A transcript-based editor is the fastest way to do this: you edit the words and the video follows. Descript works this way, adds captions, and cleans audio with Studio Sound. Hobbyist is $16 a month billed yearly ($24 monthly) with 1080p export and no watermark; Creator, $24 billed yearly, adds 4K and the stock library.11

  1. Import voice and clipsPut the narration on the timeline first, then drop each scene under its line.
  2. Trim to the voiceCut every clip so it changes on the sentence, not halfway through a word.
  3. Add on-screen text and captionsFive words or fewer per card; burn in captions for social feeds that autoplay without sound.
  4. Add music lowKeep it well under the voice, and check the license.
  5. Export each size16:9 for YouTube and websites, 9:16 for Reels, Shorts and TikTok.

Tip: Watch the export once with the sound off and once with your eyes closed. Each pass catches a different kind of mistake.

Tools for each step

Starting prices per month, US, checked on vendor pages on 23 Oct 2026.
StepToolStarts atFree option
One-click from documentsGemini Notebook Video OverviewsFree with standard limitsYes
Script and scene planClaude$17 Pro, billed yearlyYes
AI scenes, many modelsFLORA$18 per seatImages only
AI scenesKling$8.80 after first monthCheck Kling's site
Screen recordingTella or Cap$13 Tella; $8.16 Cap ProCap, personal use
Motion graphicsHyperFrames with Claude CodeFree framework; Claude Code needs a paid planFramework only
VoiceElevenLabs or Murf$6 ElevenLabs Starter; $19 Murf CreatorYes, no commercial use
Talking presenterHedra$20 BasicNot listed
Edit and captionsDescript$16 Hobbyist, billed yearly720p with watermark

Turning the same idea into an ad rather than an explainer? See how to make UGC video ads with AI. For a deck version, see the best AI presentation tools.

How we built this guide

The workflow comes from explainer-video methods we have covered in the Uncover AI newsletter during 2026, rewritten as one sequence. Every price, plan and feature was checked on the vendor's own pricing, help or GitHub page on 23 October 2026. We have not benchmarked the video models against each other, so this guide does not rank them on quality.

Questions people ask

Can AI make an explainer video for free?

Yes, with limits. Gemini Notebook makes Video Overviews from your documents on its free tier, and ElevenLabs and Murf have free voice plans, but those free voice plans do not allow commercial use.

What is the best AI explainer video generator?

For a video built from documents, Gemini Notebook's Video Overviews need no editing. For a branded video, there is no single tool: a chat assistant for the script, a video model or screen recorder for visuals, an AI voice and an editor like Descript give you control at each step.

How long should an explainer video be?

Most work at 60 to 90 seconds: about 140 to 160 spoken words per minute. Make a separate 9:16 cut for social if the main video is longer.

Can I use AI voices in a commercial explainer?

On paid plans, yes. ElevenLabs includes a commercial license from Starter at $6 a month, and Murf includes commercial rights from Creator. Their free plans do not.

About the authorFounder, Uncover AI

Anthony Chebat is the founder of Uncover AI and writes the Uncovering AI newsletter, read by more than 195,000 people. He has written about AI tools in the newsletter since 2023. Every guide here is checked against the vendors' own pricing and help pages and updated when plans or models change.

Anthony on LinkedIn

Sources

  1. Gemini Notebook Help: Video Overviews
  2. Claude pricing
  3. FLORA pricing
  4. Kling AI (Kling 4.0)
  5. Kling AI membership plans
  6. Google One AI plans
  7. Gemini subscriptions (Flow credits)
  8. Higgsfield Help: connect Higgsfield to an AI agent
  9. ElevenLabs pricing
  10. Murf pricing (Studio)
  11. Descript pricing
  12. Tella pricing
  13. Cap pricing
  14. HyperFrames on GitHub
  15. Hedra Play pricing
  16. Hedra Character 3 model page