How to Create AI Videos: A 30-Second Project, Step by Step

Difficulty: Beginner | Estimated time: 60–90 minutes for your first video

---

To create an AI video that looks intentional, write the script and shot list first, generate one short clip per shot (starting from a still image when you need the same product or character in several shots), then assemble, voice, caption and export in an editor. The generator makes the pictures; the script, the order and the finishing make it a video.

This tutorial builds one concrete project — a 30-second teaser for a reusable water bottle — so every step has something real to do. Swap in your own product or topic as you go.

What you need

• A generator that does text-to-video and image-to-video. We use PixVerse as the example: its site describes it as "Prompt or image in. Video out." and lists web, iOS and Android apps.
• An editor with captions and audio tools. Filmora lists Auto Caption and AI Audio Denoise among its features; any timeline editor works.
• A voice-over — your own recording or a text-to-speech tool.
• About 60–90 minutes and some generation credits. Check your plan's allowance before you start.

Step 1: Write a 30-second script

Thirty seconds of narration is short — around 75 words at a relaxed speaking pace. Read it aloud with a timer and cut until it fits.

Example script:

"Most bottles leak, sweat or smell after a week. (pause) This one doesn't. Double walls keep drinks cold for your whole day. One hand opens it, and the lid locks shut in your bag. It fits a car cup holder and a bike cage. Fill it once, forget about it. Find it at our shop today."

Step 2: Turn the script into a shot list

Give every sentence a shot. Keep each shot at about five seconds — short clips are cheaper to regenerate when one goes wrong.

Shot 6 is text and logo — make it in the editor, not the generator, so the spelling is exact.

Step 3: Make one reference image of the product

When the same object appears in several shots, generate (or photograph) one clean still first and animate from it, rather than describing the bottle from scratch five times. PixVerse lists image-to-video alongside text-to-video, and its site also offers GPT Image 2.5 for creating stills in the same account. A real product photo is better still: it is your actual product.

Step 4: Write each prompt with the same six parts

Use one template for every shot so the style stays consistent:

Subject + action + setting + camera + lighting/style + length.

Shot 2 example: "A matte green insulated water bottle standing on a wooden desk, dry surface with no condensation, slow camera push-in, soft morning window light, clean commercial style, 5 seconds."

Keep the lighting/style words identical across all shots. Change only subject, action and camera.

Step 5: Generate, review, regenerate

• Generate shots 1–5, then watch each one at reduced playback speed. Look for warped hands, flickering edges, objects changing shape and any text that appears in the frame.
• Regenerate only the failing shots. Change one thing at a time — usually the camera move or the action — so you learn what fixed it.
• Expect two or three attempts on the hardest shot (hands are the usual culprit in shot 3).

Step 6: Assemble on a timeline

Import the five clips into your editor in shot-list order and trim each to its five seconds. Build shot 6 with a text layer and your logo. Add a short cross-fade only where a cut feels abrupt.

Step 7: Add the voice-over and music

Record the script or generate it with a text-to-speech tool, then line each sentence up with its shot. Run noise reduction on a home recording (Filmora, for example, lists AI Audio Denoise). Add music well below the voice — if you have to strain to hear the words, it is too loud.

Step 8: Captions, check list and export

Generate captions automatically (Filmora lists Auto Caption), then read every line against your script and fix brand names and numbers by hand. Before exporting, check:

• Product name, price and any claims on screen are exactly right.
• No generated text or logos you do not own appear in any frame.
• The first three seconds work with the sound off.
• The last frame shows where to buy.

Export vertical (9:16) for short-form social platforms and horizontal (16:9) for YouTube and your website. If you use Filmora's free version, note that its exports carry a watermark.

Troubleshooting

Frequently asked questions

How long does it take to make an AI video?

A first 30-second video typically takes an afternoon including retries; with a script template and shot-list habit, later ones go much faster. Generation itself is the quick part — writing, reviewing and finishing take most of the time.

How do I keep the same character or product across AI video shots?

Start each shot from the same reference image using image-to-video, and reuse the same style words in every prompt. Some generators, PixVerse among them, also list character-consistency features.

Can I add text and logos in the AI generator?

You can try, but generated lettering is unreliable. Add titles, prices and logos as layers in your editor so they are exactly right.

Which AI video tool should a beginner use?

A generator with both text-to-video and image-to-video, plus any editor with automatic captions. Our best AI video generators of 2026 compares PixVerse, InVideo, DomoAI and Filmora from their vendor documentation.

Keep going

For script-to-video editing with AI agents, look at InVideo, which positions itself as an agent-driven editor. The finishing tools are compared in AI video editing software, licensing for your visuals in AI images vs stock templates, and better prompts in our tutorial on writing better AI prompts.

Sources checked 16 October 2026: pixverse.ai; filmora.wondershare.com; invideo.io. Editorial tutorial.