On this page
- The steps
- Real makeup video ads
- Pick one idea and one viewer
- Write the script with AI, then cut it
- Match each scene to a visual
- Record the voiceover and edit to it
- Add captions and set the format
- What long-running ads teach explainers
- Prompts to copy
- A 60-second explainer, scene by scene
- Mistakes to avoid
- Questions
Real makeup video ads to model yours on
3,632 video ads from 365 advertisers in the Meta Ad Library, read October 5, 2026. The median one ran 22 days, and an ad still paid for after months is the closest public sign that it works. Live ads first.
- Ran under a day
- Ran 7 days
- Ran 2 days
- Ran 7 days
- Ran under a day
- Ran 4 days
- Ran under a day
+3,620 more makeup video ads
Hooks, landing pages and run length for each
- Ran under a day
- Ran 6 days
- Ran 6 days
- Ran 4 days
Every creative, with its copy, format and run length
The openers that ran longest
40% OFF + FREE Tint – Lashes You’ll LOVE!
Luxe Cosmetics · ran 504 daysChat With Us
Isabella Garcia International · ran 455 daysRealize Sonhos, Seja Consultora
Natura · running 434 daysBrushes Are Back in Stock
Morphe · ran 429 days
Pick one idea and one viewer
An explainer fails when it tries to teach three things. Before you open any tool, finish this sentence: “After watching, [audience] will understand [one idea or task].” A makeup brand might write: “After watching, women over 50 will understand why a tinted SPF replaces foundation and sunscreen in one step.” That single sentence decides your script, your visuals and your first line.
Decide where the video will run. A 60-second cut suits a landing page or a YouTube embed. A 15 to 30 second cut suits paid social. Write the longer version first, then trim. Every scene you cut from the long version is one less scene to generate, check and fix.
Gather your inputs now. You want product photos on clean backgrounds, a screen recording if you sell software, your brand colors and fonts, and one proof point you can actually back up. Do not ask AI to supply statistics. It will happily invent them, and you will be the one who has to defend them.
Write the script with AI, then cut it
Use AI for the first draft, not the final one. Paste the script prompt below and fill in the topic, audience and goal. Ask for 5 to 7 scenes, each with narration, what to show and any short on-screen text. The shape that works is simple: a relatable problem, the idea in plain language, one concrete example, and a single takeaway.
Read the draft out loud. Anything that sounds stiff in your mouth will sound worse in a voiceover, so rewrite it. A comfortable speaking pace is roughly 130 to 150 words a minute, so a 60-second explainer is about 130 to 150 words and a 30-second cut is about 70 to 75. If you are over, cut scenes, not syllables.
Then fact-check line by line. Delete every number you cannot source and every promise you cannot substantiate. Replace vague claims with something specific the viewer can picture, such as “one step instead of two.” Keep the script in a plain document, one scene per row, so you can regenerate a single scene later without touching the rest.
- One idea per video, one sentence per beat
- Problem first, takeaway last
- Write for the ear, not the page
- No invented facts or statistics
Want makeup video ads like these for your product?
The makeup video ads that kept running, read from the ad libraries. Every change waits for your approval.
Match each scene to a visual
The rule is that visuals prove the narration. If you explain software, record the real screen and zoom into the clicks that matter. If you explain a concept, use simple diagrams, icons or short AI clips. For a physical product, feed your own product photo into an image-to-video tool so the packaging and shade stay accurate instead of being reinvented.
Generate one scene at a time. Keep AI clips short, around 3 to 5 seconds, and build a 6 to 10 second scene from two or three of them. In each prompt describe one subject, one camera move and one lighting setup. Leave text out of the prompt. AI video still garbles lettering, so add titles and labels in your editor where you control spelling.
Keep the look consistent. Write a one-line style note, for example “soft daylight, cream backdrop, muted rose and white palette, vertical 9:16,” and paste it into every prompt. Make two or three takes per scene and keep the best. Mixing a real screen recording with AI clips is fine as long as the color and pacing stay in the same family.
Record the voiceover and edit to it
Narration goes on the timeline first. Use your own voice if you can, because a slightly imperfect human read often feels more trustworthy than a polished synthetic one. If you use text-to-speech, listen to the whole track and respell any product names it mispronounces. For a creator-style delivery, see how UGC ads handle a spoken, to-camera tone.
Now cut visuals to the sentences. Change something on screen every 3 to 5 seconds: a new clip, a zoom, a label appearing. Land each visual a beat before the word it illustrates, so the viewer sees the idea as they hear it. Trim silence between sentences until the pace feels slightly brisk.
Add music last and keep it low enough that the voice stays clear on phone speakers. Export a draft and check three things: does every scene show what is being said, does anything feel padded, and could you delete the weakest scene without losing the point? If yes to the last one, delete it.
Add captions and set the format
Build vertical first. Use 9:16 at 1080 by 1920 for Reels, TikTok and Shorts. Place captions in the middle of the frame, away from the bottom and right edges where buttons, handles and descriptions cover the picture. Exact safe zones differ by placement, so check the Meta ad specs before you finalize a paid version.
Burn captions into the video. Many people watch with the sound off, so the captions have to carry the explanation alone. Use 3 to 6 words per line, high contrast, one font, and highlight the keyword if your editor supports it. Proofread the auto-generated text, because names and product terms are where captions go wrong.
Finish with one action and one frame. The last 2 to 3 seconds should hold a single call to action over a clean background, with the product or screen still visible. Then preview on an actual phone, at arm’s length. If you have to squint, the text is too small.
What long-running ads teach explainers
The Sprites ads library shows real ads from a category. In makeup it covers 3632 video ads from 361 advertisers, with a median run of 21 days. The ads that ran far beyond that open by naming the promise or the viewer in the first line. Sheryl Davis ran “Flawless Skin In Seconds” for 294 days. Vibriance ran “Finally, Tinted SPF Made For Women 50+” for 280 days. Wonderskin ran “The Blush That Becomes Uniquely Yours” for 185 days, and Lawless ran “Ditch the Needles - Shop Lawless” for 184 days.
The pattern is one clear claim, said immediately. Vibriance names the audience and the product type in one line. Lawless names what you get to skip. Sheryl Davis names the speed of the result. None of them warm up. Your explainer’s first line should do the same: state the problem or the payoff, then spend the rest of the video showing how. See more of this in makeup video ads and the wider video ads guide.
Use that for iteration. Keep the body of your explainer and make two or three versions of only the first 3 seconds: one that opens on the problem, one on the payoff, one on the audience. Run them against each other and keep the winner. If you would rather not produce the variants yourself, Director by Sprites is “a full creative team for your startup.” Every Monday you get finished, on-brand video ads, reviewed by a human.
Prompts to copy
Paste each one as written, then change the product, the setting and the light to match your own footage.
Script: scene-by-scene draft
Create a 60-second explainer video script about [topic] for [audience]. By the end, viewers should understand [goal]. Use plain language and one concrete example. Break the script into 5 to 7 scenes. For each scene, give the narration, what to show, and any short on-screen text. Open with a relatable problem and end with one clear takeaway. Do not invent facts or statistics.
Hook: the problem scene
Vertical 9:16, 4 seconds. A cluttered bathroom shelf with several small jars and tubes, seen from slightly above, soft daylight, cream and rose palette. The camera slowly pushes in as one tube slides forward out of the clutter. Calm, clean, editorial look. No people, no logos, no readable text on any packaging.
Explain: the idea in motion
Vertical 9:16, 5 seconds. Minimal flat illustration on a cream background: two simple circles labeled by color only, one pale yellow and one soft pink, slide together and merge into a single warm peach circle. Smooth easing, gentle camera hold, clean vector style, rose and white palette. No letters, numbers or logos in the frame.
Example: product close-up
Vertical 9:16, 5 seconds. Macro shot of a plain unbranded tinted lotion tube on a pale marble surface. A small swirl of cream squeezes out and settles into a smooth curl. Soft side light, shallow depth of field, slow push-in. Warm neutral tones, premium and quiet. No hands, no people, no text or logos.
Proof: one step instead of two
Vertical 9:16, 5 seconds. Top-down view on a cream surface: two unbranded bottles on the left fade out while one single unbranded tube on the right stays and rotates slowly into center frame. Soft shadows, even daylight, clean minimal styling, muted rose palette. No people, no brand marks, no readable text.
Takeaway: clean end card background
Vertical 9:16, 3 seconds. A calm cream-to-blush gradient background with soft grain and a gentle slow drift of light, leaving the center and lower third empty for text added later. Subtle, premium, minimal. No objects, no people, no logos, no text generated in the frame.
A 60-second explainer, scene by scene
| Scene | Seconds | Narration job | Visual |
|---|---|---|---|
| 1. Hook | 0 to 5 | Name the problem or payoff in one line | Relatable AI clip or product shot |
| 2. Problem | 5 to 15 | Show why the old way is annoying | Cluttered or confusing scene, real screen if software |
| 3. Idea | 15 to 30 | Explain the one idea in plain words | Simple diagram or animated icons |
| 4. Example | 30 to 45 | Walk through one concrete case | Product close-up or screen recording |
| 5. Takeaway | 45 to 55 | Repeat the idea in one sentence | Clean recap frame with short text |
| 6. Action | 55 to 60 | Say the one thing to do next | End card with button text |
Mistakes to avoid
- Trying to explain two or three ideas in one video instead of one.
- Generating the whole video in a single prompt instead of scene by scene.
- Asking the AI to render text, labels or logos inside the clip instead of adding them in the editor.
- Letting the model invent statistics or claims that you never checked.
- Editing visuals first and recording narration after, which forces awkward pacing.
- Placing captions in the bottom corner where platform buttons and descriptions cover them.
Rather have the videos made for you?
Director by Sprites is a full creative team for your startup. Every Monday you get finished, on-brand video ads, reviewed by a human. You never write a prompt.
Apply to Director by Sprites →Questions
What is the easiest way to make an explainer video with AI?
Write a short script, split it into 5 to 7 scenes, and use AI to help with visuals, narration and captions. For a first video, a voiceover over screenshots or simple slides is enough. You do not need AI-generated footage for every scene.
How long should an AI explainer video be?
Around 60 seconds works for a landing page or YouTube. For paid social, cut a 15 to 30 second version from it. Write the long script first, then remove scenes until only the one idea remains.
Can AI write the script for my explainer video?
Yes, for the first draft. Give it your topic, audience and goal, and ask for 5 to 7 scenes with narration, visuals and on-screen text. Then read it aloud, cut stiff lines and delete any fact or number you cannot verify.
Should I use AI voiceover or my own voice?
Either works. Your own voice often sounds more trustworthy, while text-to-speech is faster to revise. If you use text-to-speech, listen to the full track and fix any mispronounced product names before editing.
What aspect ratio should an AI explainer video be?
Use 9:16 at 1080 by 1920 for Reels, TikTok and Shorts. Keep captions and key visuals in the middle of the frame so buttons and descriptions do not cover them. Use 16:9 only if the video will live on a website or in a desktop player.
How do I make my AI explainer video better after the first version?
Keep the body and change only the opening. Make two or three versions of the first 3 seconds, one on the problem, one on the payoff, one naming the audience, and compare them. Long-running ads in the category tend to state a clear claim in the very first line.