How to Make YouTube Shorts With AI

To make YouTube Shorts with AI, write the hook first, keep the script to one point and about 150 words or fewer, generate it natively in vertical 9:16 with AI narration, word-by-word captions and music, check the captions sit clear of YouTube's on-screen controls, then publish several variations and keep the shape that viewers watch through.

Shorts rewards a specific set of behaviors, and most of them have little to do with production quality. This is a process for making Shorts with Vidcru that respects what the format actually optimizes for.

What Shorts is measuring

Before the craft, the mechanics. Shorts distribution leans heavily on viewer behavior:

• Whether people stay past the opening seconds. If viewers swipe away immediately, the video stops being shown to more of them.
• Watch-through and rewatches. A short video that people finish, or loop, sends a stronger signal than one they abandon halfway.
• Consistent posting. The system needs repeated signal to work out who your audience is.

Notice what is missing: resolution, cinematography, edit polish. Those affect whether people subscribe, not whether the video gets shown. Optimize for being watched first.

How to make a YouTube Short with AI, step by step

1. Write the hook before anything else

The first line does more work than the rest of the video combined. Write it first, in isolation, and write several.

Hooks that work fall into a few shapes:

• The claim. "You are using AI video generators wrong."
• The question with a stake. "Why do your Shorts stop at 200 views?"
• The result first. "This took four minutes to make."
• The contradiction. "Longer videos are easier to rank than Shorts."

Hooks that do not work: introductions, context-setting, "in this video I'm going to". There is no budget for a preamble. Start at the point.

2. Keep the script to the length

A 60-second Short is roughly 130 to 150 spoken words. Most drafts are two to three times that, which produces narration so rushed the captions cannot keep up.

Write the script, count the words, and cut to the number. This is the least glamorous step and the one that most improves the finished video.

One point per Short. If you have three points, that is three Shorts, and three focused Shorts usually do better than one video that tries to hold all three.

YouTube has extended the maximum Short length beyond 60 seconds for many uploads, but being allowed to go longer is not a reason to. Vidcru's Agentic Video Shorts templates run up to 90 seconds, which is plenty for one well-made point.

3. Pick the generation method to match the content

This is where AI tooling gives you a real choice, and the wrong choice is a common reason a Short underperforms.

Story, history, facts. Use AI Images in 9:16: a newly generated image for every scene, with camera movement, in the art direction you choose. This is the format behind a lot of the story content in the Shorts feed. Moving AI images explains how camera motion on a still works and when it is the better choice.

A repeatable series: lists, how-tos, comparisons, explainers. Use an Agentic Video Shorts template. The template fixes the look, so every Short in the series is recognizably yours, and the agent builds each new script into it with scenes, voice and music.

Anything advice-shaped, such as finance, health or education, or direct address to camera. Record yourself and upload the recording to Edit Your Video, which cuts, captions and scores it for you. A real person reads as more credible, and in categories where trust drives the follow, that matters more than anything a generator can add.

Agentic Character Video, where your own character stars in every scene, is widescreen 16:9 only. It belongs on your main channel rather than in Shorts.

4. Choose the voice and caption style

Pick one of 30 AI voices; narration is available in many languages. Keep the same voice across a series so returning viewers recognize it. Then choose a caption style. Captions are timed word by word to the narration.

5. Get the vertical framing right

Do not crop a widescreen video to 9:16. Cropping throws away the sides of every frame, and the subject is often in the part you threw away. Generating natively vertical means the composition was built for a phone.

Two things to check on every Short:

• Captions inside the safe zone. YouTube overlays a title, channel name and action buttons on the video. Text underneath any of that is text nobody reads. Keep captions in the middle band.
• Subject sized for a small screen. What reads on a monitor can be illegible on a phone held at arm's length. When in doubt, go closer.

6. Treat captions as the main text

Many people scroll Shorts with the sound low or off, and in those sessions the caption carries the content. That makes captions a primary element rather than an accessibility extra. Vidcru adds word-by-word captions to every method; whichever you use, watch the preview and check the placement before exporting.

7. Export and upload

Download the MP4 in Full HD 1080p and upload it through the YouTube app or YouTube Studio. Vidcru does not publish to YouTube for you, so you write the title, description and any hashtags yourself. Finished videos stay in your History for 30 days if you need to download them again.

Publish enough to learn something

This is where most people fail, and it is not a craft problem.

Which hook works is genuinely hard to predict. The only reliable method is to publish several and read the retention data in YouTube Studio. That means your production process has to support volume, which is the practical argument for generating rather than filming: the cost of a fifth variation is credits, not another afternoon.

A workable cadence:

1. Write five hooks against one idea.
2. Generate all five with the same body and template.
3. Post them across a week.
4. Keep the shape that retained best; discard the rest.
5. Write five variations of the winner.

Should you make Shorts and long-form videos together?

If you run a channel, yes, but make each in its own format. Vidcru's Agentic Video has a long-form mode in 16:9 for videos up to 15 minutes and a Shorts mode in 9:16 for up to 90 seconds. Write the long video first, pull the single strongest point out of it, and generate that as its own vertical Short rather than cutting the widescreen version down. How to start a faceless YouTube channel with AI covers how the two fit into one publishing schedule.

What to check before posting

• Does the first line work with no context?
• Is the script under 150 words?
• Are captions clear of the top and bottom interface zones?
• Does it make exactly one point?
• Would it still make sense with the sound off?

Common questions

Can you make YouTube Shorts with AI without showing your face?

Yes. AI Images and Agentic Video Shorts produce a complete narrated video with no one on camera, which is how most faceless Shorts channels work.

Can I monetize AI-generated Shorts?

YouTube's policies focus on whether content is original and adds value, not on which tool made it, and they change over time. Check the current YouTube Partner Program rules, and avoid mass-producing near-identical videos.

How much does a YouTube Short cost to make?

Shorts are the cheapest videos to make because they are short. Vidcru sells one-time credit packs with no subscription, valid for 90 days, and every video shows its credit estimate before you create it. See pricing for current packs.

Start your first Short

Open Agentic Video Shorts, paste a script of 150 words or fewer, pick a voice and caption style, and preview the result before you export.

For the full comparison of methods, read the AI video generator guide. If TikTok is your other platform, how to make TikTok videos with AI covers what changes there.