How to make a long-form AI video

Why most AI video generators stop at seconds

Text-to-video models generate a short clip at a time, usually a few seconds and rarely more than a minute. An hour of video made that way means hundreds of separate clips, each priced as video, each drifting a little in how characters and places look, and all of them to line up in an editor afterwards.

Long narrated videos rarely need full motion in every shot anyway. Documentaries and video essays have long relied on still images with slow camera moves. So the tools that handle long videos well treat the narration as the spine and make one picture per scene, with motion added on top.

What to look for in a long-form tool

• A per-project length limit that covers your longest video, stated plainly.
• The whole video in one project and one timeline, with nothing to stitch afterwards.
• One visual style held across every scene, and recurring characters that stay recognizable.
• Changing one scene without regenerating the rest.
• The cost shown before each step, since an hour of scenes adds up.
• Processing that keeps going after you close the tab.
• Chapters, subtitles and an .srt file, and commercial rights if you'll monetize.

The steps, start to finish

1. Plan the idea and the length

   Pick a topic with enough material for the length. Narration runs at roughly 130 to 160 words a minute, so a 20-minute video is about 3,000 words and an hour is around 9,000. Outline it in parts first: an hour-long history is easier to write as six 10-minute chapters than as one block.

2. Write or draft the script

   Write for the ear: short sentences, one idea at a time, and a clear signal when the subject changes, because that's where the picture will change too. If an AI drafts it, read every line and check every fact. A wrong date in minute 40 is still wrong.

3. Choose the narration

   Your own recording gives the channel a voice people recognize. An AI voice is quicker to redo when you change a line. Either way, use the same voice on every episode, and listen to a sample at the speed you'll use before committing to an hour of it.

4. Pick one visual style and keep it

   Mixing styles in one video looks like a mistake. Choose one look that suits the subject, such as archival photographs for history or a painted style for fiction, and hold it for the whole video and ideally the whole channel. Keep recurring people and places described the same way each time they appear.

5. Set the scene pacing

   One picture every 6 to 12 seconds suits most narration. Change it faster in tense or busy stretches and slower in reflective ones; sleep videos can hold a picture for much longer. Slow pans and zooms keep a still from feeling static.

6. Add chapters, music and subtitles

   Chapter cards between parts give viewers a place to rest and rejoin. Keep music well under the voice. Subtitles help the many people who watch with the sound off, and an .srt file uploaded to YouTube gives you accurate captions there too.

7. Export and upload

   Export at 1080p in 16:9. In the YouTube description, list chapter timestamps starting at 0:00 so the video gets chapters on its progress bar. Write a title and thumbnail that promise what the video actually delivers.

Making it in Animatic

Animatic is built around these steps, and it keeps the whole video in one project however long it runs.

1. Start. Upload your own recording with Start with Narration, paste a script and pick from 62 AI voices in 15 languages with Start with a Script, or use Start with an Idea to get an editable script drafted at the length you ask for.
2. Plan. Animatic transcribes the narration and splits it into scenes where the subject changes. You choose how often the picture changes.
3. Style. Choose one of 20 built-in Visual Styles, or on a paid plan save your own from a prompt and up to 3 reference images. Continuity keeps recurring characters and places looking the same.
4. Generate. Make my video plans the scenes and makes every picture in one go, after you confirm the credit cost. It keeps running after you close the browser.
5. Edit. Remake or rewrite any single scene, change its camera move and transition, and in AI-voiced projects change one scene's line and only that part is re-voiced. Animate scene turns chosen pictures into short video clips.
6. Finish. Chapter cards and sound effects come with every plan. Paid plans add music (a library of 80 tracks, or automatic music mixed under the narration) and burned-in subtitles, plus the .srt, script and MP3 downloads.
7. Export. 16:9, 9:16 or 1:1. Paid plans export 1080p with no watermark and include commercial use.

How long a project can be

On Free, AI-written scripts and AI narration stop at 5 minutes, and exports are 720p with a watermark. The length is a limit per project, not an amount of video your credits are guaranteed to cover.

What an hour costs in credits

Each scene picture is 3 credits at Standard quality, and AI narration is about 3 credits a minute. An hour with a new picture every 10 seconds is 360 scenes, so roughly 1,080 credits for pictures and 180 for narration. With your own recording, the narration costs nothing. Cut rhythm can cut between wide, medium and close framings inside a long scene, so you can use fewer pictures without the video feeling slower. Studio shows the exact cost before every step, so these numbers are a guide, not a quote.

YouTube rules are yours to meet

YouTube asks creators to disclose realistic altered or synthetic content when they upload, and its monetisation policies exclude content that's mass-produced or repetitive. Whether a video qualifies depends on what you put into it: your research, your script, your editing.

If you're planning a video and want to check the pacing before final pictures, see what an animatic is. For what Animatic offers faceless channels in particular, see the faceless YouTube video generator page.

Questions