How to Make Long Documentaries With the CapCut AI Video Maker

CapCut's AI Video Maker can turn a written script into a narrated documentary with matched stock footage, but each project only takes about 3,000 words. For a long documentary, you write the full script first, feed it to CapCut in sections, export each one and join the exports on one timeline.

This workflow suits calm, long-form formats like sleep documentaries, history, science, mythology and biographies, where the story matters more than fast editing.

How to make a long documentary in CapCut

1. Write the full script chapter by chapter. Have the AI write a detailed chapter outline first. Then write one chapter at a time, and before each new chapter give it a short summary of everything written so far. That way chapter 8 still knows what happened in chapters 1 to 7.
2. Split the script into sections under 3,000 words. Cut at chapter breaks so each section starts and ends at a natural point.
3. Paste the first section into AI Video Maker. Open CapCut, start the AI Video Maker from the home screen and paste in the first section.
4. Pick a calm, slow voice. For sleep and other relaxed content, the voice matters more than any editing. Pick a low-energy voice and listen to a sample before you commit.
5. Generate the video. CapCut writes the voiceover and matches a stock clip to each paragraph. Expect about 90% of the result to be usable as it is.
6. Fix the clips that don't fit. Play it through and replace any stock footage that doesn't match what the narrator is talking about.
7. Export the section. Export it as its own video file and give it a numbered name so the order stays clear.
8. Repeat for every section. Run steps 3 to 7 for the rest of the script. Use the same voice each time so the narration sounds like one person from start to finish.
9. Combine the exports and render. Start a new project, drag every section export onto one timeline in order and export the full documentary.

Write the script with the API instead of the chat

The script holds viewers far more than the visuals do. If it repeats itself, drifts off topic or loses momentum halfway, people stop listening, and better footage won't win them back.

The chat window makes long scripts hard. By chapter 4 you're typing "continue" and "don't repeat what you already said", and the model starts contradicting earlier chapters.

Running the same chapter-by-chapter loop through the Claude API fixes that. A small tool, like a Google Sheet connected to the API, sends each prompt for you, passes along the summary of earlier chapters and collects the output. You don't have to babysit it, and you pay per use instead of another monthly subscription.

For more on writing lots of scripts without them sounding the same, see how to write scripts in batches without sounding robotic. If you're still deciding which parts of the job to hand to AI, what AI video tools can and can't do yet covers that.

FAQ

Why does CapCut cut off my documentary script? The AI Video Maker takes roughly 3,000 words per project. A 90 minute documentary runs 15,000 to 20,000 words, so you have to split it into sections, export each one and join them afterwards.

How much does a long AI documentary cost to make this way? A 20,000 word script costs around 35 cents through the Claude API, and a CapCut subscription is about $20 a month with unlimited exports. At 30 to 40 videos a month that works out to roughly$1 per video, not counting research, thumbnails or your time.

Will YouTube flag these videos as reused content? Real stock and historical footage mixed in with AI assets is your best protection against reused or inauthentic content flags. You still need titles and thumbnails that don't repeat from video to video.