How to make a 60-second AI explainer that doesn't look AI
A repeatable workflow for explainer videos with real motion, real voiceover, and zero uncanny-valley moments. Used by founders to ship product launches in a single afternoon.
Most AI explainer videos look like AI explainer videos. The voiceover sounds slightly off, the visuals don't quite match the words, and the whole thing has that "made on a Sunday with no editing" quality. That's not an AI problem, that's a process problem.
This guide gives you the exact 60-second explainer workflow that founders use to ship product launches and that works for B2B SaaS, courses, consumer apps, and physical products alike.
The three-tool stack
You don't need After Effects, Premiere, or any traditional editing skills.
- Runway Gen-4.5 for short generative motion clips (logos materializing, abstract data flowing, product silhouettes spinning).
- ElevenLabs for the voiceover, v3 voices only.
- CapCut for the edit, beat-sync, and final polish.
Step 1: Write the 60-second script (15 minutes)
A 60-second explainer is exactly 150-160 words at a natural pace. Not 200. Not 250. Cutting from 250 to 160 is where most founders refuse to do the work, and it's what kills the video.
Structure that works:
Write it. Read it aloud. Time yourself. Cut every "really," "very," and "amazing." If you're over 65 seconds at a natural pace, cut more.
- Seconds 0-5: The problem. One sentence. Visceral. "If you've ever spent a Sunday rebuilding the same spreadsheet, this is for you."
- Seconds 5-25: The solution. Three sentences. Concrete. What it is, who it's for, why it's different.
- Seconds 25-50: One concrete result. A number, a customer quote, a before/after. The thing that makes it real.
- Seconds 50-60: The CTA. One ask. "Try it free at [domain]." Repeat the domain twice if you can.
Step 2: Storyboard the visuals (10 minutes)
For every sentence, write what's on screen. Don't skip this, the temptation is to "figure out visuals while editing," and that's how you end up with random stock footage that doesn't connect to anything.
Format:
- "If you've ever spent a Sunday rebuilding the same spreadsheet…" → animated spreadsheet cells filling and emptying, dark UI
- "This is for you." → cut to product logo, hold 1 sec
- "[Product] is [category] for [user]." → product UI screen recording, fast scroll
- … etc.
Step 3: Generate the voiceover in ElevenLabs (5 minutes)
Open ElevenLabs, pick a v3 voice. The 2026 v3 library has voices that sound like real people doing real reads, pacing changes, breath sounds, slight emphasis variation. Avoid the older Multilingual v2 voices for explainer work; they sound flat next to v3.
Settings for explainers:
Generate the full script as one take. If a single sentence reads wrong, regenerate just that sentence using the v3 sentence-level regen. Don't redo the whole thing, voice consistency matters more than fixing one okay-but-not-perfect line.
- Stability: 40 (some natural variation, not too unstable)
- Similarity: 75
- Style: 30-40 (matters for explainers, punctuation actually affects delivery)
- Speaker boost: on
Step 4: Generate motion clips in Runway (30 minutes)
For every "abstract motion" beat in your storyboard (~5-8 clips for a 60-second video), generate a 4-5 second clip in Runway Gen-4.5.
Prompts that work for explainers:
Use reference images from your brand for consistency. If you have a logo, drag it into Runway as a reference and prompt for it to materialize. If you have a screenshot of your UI, reference it and prompt for an animated version.
For 5-8 clips, expect to generate 15-25 candidates and pick the best of each. AI video isn't deterministic; expect to roll the dice 3-4x per beat to get one that lands.
- "Abstract data points flowing from left to right across a dark surface, soft glow, minimal, 4 seconds"
- "A logo emerging from particles of light, dark background, premium tech feel, 3 seconds"
- "Top-down view of a clean workspace, papers being organized into neat stacks by invisible hands, soft daylight, 5 seconds"
Step 5: Edit in CapCut (20 minutes)
This is where most explainers come together, or fall apart. The order of operations matters:
Export at 1080x1920 (vertical), 1920x1080 (landscape), or 1080x1080 (square) depending on where you'll post. For best results, export the master in landscape and re-frame in CapCut for vertical/square versions.
- Drop the voiceover on track 1 first. Always. Voice dictates the cut.
- Place each visual where its sentence starts. Use markers on the timeline; CapCut shows them clearly.
- For UI screen recordings, speed them up to 1.5x or 2x. Real UI footage is too slow for a 60-second cut.
- Add background music at -18 dB so the voiceover sits clearly on top. CapCut's library has thousands of free royalty-cleared tracks; pick something cinematic in 100-115 BPM range.
- Beat-sync the cuts to the music using CapCut's auto-beat feature. Visual changes on beats feel professional. Visual changes between beats feel amateur.
- Add subtle motion to static screenshots, a slow 5-second zoom, alternating direction across shots. Static UI screenshots in motion video feel dead.
- Title card at start (0.5 sec) and end (1 sec). Just your logo and the URL. Don't over-design these.
What this looks like in practice
The first explainer takes 3 hours. The fifth takes 60 minutes flat.
Mistakes that kill explainers
- AI generated humans. Even in 2026, AI humans in explainer roles trip the uncanny valley for most viewers. Use abstract motion, product UI, or real human B-roll from a stock library, not generated faces.
- Music louder than voice. Single biggest reason a video feels amateur. -18 dB music, voice at 0 dB, never the other way around.
- No CTA. A polished explainer with no clear ask leaves viewers without a next step. Always end with one specific action.
- Vertical-first thinking on a desktop product. If your CTA is "go to [domain].com on your laptop," vertical isn't the right primary format. Make a 16:9 master, post a vertical cut as a teaser.
Where to go next
After your first explainer ships, the upgrades are:
Make the 60-second version this weekend. Ship it. Watch it convert. Then make the longer cut once you know which 30 seconds were doing the actual work.
- A 30-second cut for paid social ads. Same script, halved.
- A 3-minute deep-dive for your homepage and YouTube. Same workflow, more clips, more depth.
- A localized version in 2-3 other languages using ElevenLabs Dubbing. Massive distribution lift for almost no extra work.