The most expensive resource on a Shopify product page is not traffic. It is the shopper who watches, understands, and still leaves. Video is the cheapest way to recover that person: it turns abstract spec text into a use case someone grasps in three seconds. And what actually holds them in a product video is rarely the footage — it is the narration.
This guide covers product video voiceover for Shopify stores: what each of the three video placements is for, how to rewrite listing copy into a spoken script, 15/30/60-second templates with timecodes, how to say specs out loud, how to plan multilingual markets, and a pre-launch checklist.
Three placements, three different jobs
One product, one video, everywhere is the most common waste in ecommerce video. The same clip cannot open a cold visitor and close a ready buyer.
| Placement | Length | Viewer’s state | Voiceover job | Typical failure |
|---|---|---|---|---|
| Hero media (first slide) | 15–30s | Arrived from an ad or search, deciding “is this for me” | Say what it is and what it solves within three seconds | Brand story in the first five seconds |
| Mid page (after specs) | 30–60s | Interested, verifying details | Explain material, usage, sizing, fit | Repeating the hero video |
| Page bottom / upsell area | 10–15s | About to add to cart | Remove the last objection (shipping, warranty, size) | Listing features again |
| Off-site ad creative | 6–15s | Does not know the brand | One pain point plus one visual proof | Too much information |
A useful division of labor: hero sells “what this is,” mid page sells “why it fits me,” page bottom sells “why now.” Do not reuse the copy across all three.
From listing copy to spoken script: five rewrites
Product copy is written for someone already reading. Narration is written for someone still scrolling. Different grammar.
| Listing copy | Why it fails when spoken | Narration rewrite |
|---|---|---|
| “High-density memory foam, 45D, 3-second rebound” | Spec stack; nothing is retained | “You sink in, and three seconds later it pushes back. That’s the difference.” |
| “Suitable for office, home, and car” | Three use cases means no use case | “On an office chair, eight hours a day, it doesn’t flatten.” |
| “SGS tested” | The certificate name paints no picture | “SGS tested — formaldehyde release below the limit.” |
| “100% cotton, skin-friendly, breathable” | Adjectives with no evidence | “Pure cotton. Ten washes, still no pilling.” |
| “Limited offer while stocks last” | Ad-speak, triggers defensiveness | “Three hundred units in this batch. The next one lands next quarter.” |
The principle: replace adjectives with verifiable specifics. Narration has no visual support, so abstract words carry nothing.
The first three seconds decide everything
| Hook type | Example | Best for |
|---|---|---|
| Pain first | “Does your back support collapse after two hours?” | Products solving a known irritation |
| Result first | “This chair got me through a full day without shifting position.” | Visibly effective products |
| Contrast | “The difference between this and the usual one is this cut.” | Visually distinct products |
| Number hook | “Three seconds and this fabric is dry.” | Quantifiable features |
Never open with “Welcome to our brand” or “Today I want to introduce.” That spends the most expensive three seconds on a greeting.
Three script templates with timecodes
15 seconds: hero and ads
[0-3s] Hook: one pain point or one result
"Two hours in, your lower back starts complaining."
[3-9s] What it is + one visual proof
"A 45D memory foam seat cushion. You sink in, it rebounds in three seconds."
[9-13s] One concrete difference
"Unlike regular foam, it doesn't stay dented."
[13-15s] CTA
"Scroll down for the size chart and pick yours."
30 seconds: full hero version
| Seconds | Segment | Voiceover note |
|---|---|---|
| 0–4 | Hook | Normal pace, no music, let the content hold attention |
| 4–12 | What it is | One noun plus one number, not three adjectives |
| 12–22 | How to use | Start with verbs: place, sit, adjust |
| 22–27 | Proof of difference | Comparative sentence that names the comparison |
| 27–30 | CTA | Slow down, point at the next step on the page |
60 seconds: mid page
60 seconds is not a stretched 30. It adds detail and objection handling:
| Seconds | Segment | Content |
|---|---|---|
| 0–5 | Continuation | “If you’re still here, you’re probably wondering about sizing” |
| 5–20 | Specs | Size, material, weight — one sentence each |
| 20–35 | Use case | One specific person, one specific moment |
| 35–50 | Objections | Cleaning, warranty, shipping |
| 50–60 | CTA | A single action |
Do not reintroduce the product at 60 seconds. The viewer scrolled down from the same page; repeating the intro makes the video feel padded.
Saying specs out loud: six places people stumble
| Written | Say it as | Why |
|---|---|---|
45D |
“forty-five D” | The letter is clearer than spelling out the density unit |
500ml |
“five hundred milliliters” | Never “five hundred m l,” never drop the unit |
2.4GHz |
“two point four gigahertz” | Skip the decimal and it becomes twenty-four |
IPX7 |
“I P X seven” | Letters separately, number as a word |
100% cotton |
“one hundred percent cotton” | Keep it consistent across locales |
S/M/L |
“three sizes: small, medium, large” | State the count so listeners know how many options exist |
Two more rules:
- Give every number a reference. “1.2 kilograms” is weaker than “1.2 kilograms — about two bottles of water.”
- Do not mix units in one video. All centimeters or all inches. Mixed units force mental conversion and people leave.
Pace and conversion: rhythm beats speed
| Segment | Suggested pace | Pause | Why |
|---|---|---|---|
| Hook | Normal (150–170 wpm) | 300ms at sentence end | Needs forward momentum |
| Explanation | Slightly slower (140–155 wpm) | 250ms between sentences | Lets information land |
| Specs | Slow (130–145 wpm) | 400ms between items | Numbers need processing time |
| CTA | Slowest (125–135 wpm) | 400ms before and after | Signals “you can act now” |
In Mandarin the equivalents run roughly 240–260 characters per minute at the hook down to 190–210 at the CTA. Never hold one pace throughout — that is what sounds like a spec sheet read aloud, and it is the most reliable tell of synthetic narration.
Pitch needs only two levels on a product page: steady while explaining, slightly lifted on the CTA. All-high sounds like home shopping; all-flat sounds like an audiobook.
Muted autoplay: captions are not optional
This gets missed constantly: most storefront themes and social feeds autoplay muted. The first impression is a silent frame.
| Required element | Reason | How |
|---|---|---|
| Burned-in captions | The only text channel when muted | Large, single line, word-for-word with narration |
| On-screen text in the first 3 seconds | The viewer has not finished reading yet | One sentence saying what the product is |
| An obvious unmute affordance | Some viewers will listen | Visible volume control |
| A soft caption track | For viewers who need to resize or switch language | Separate version for the main player |
Practical approach: export two versions — burned-in captions for the hero slot and social, soft captions for the main product player. The full accessibility treatment is in accessible e-learning voiceover.
Multilingual markets: planning voiceover across locales
Selling into five markets does not require five voices. Choose the strategy first, then the count.
| Strategy | Approach | Cost | When |
|---|---|---|---|
| One voice, many languages | One voice model generates each locale | Lowest | Brand tone matters more than local feel |
| One voice per language | Mandarin voice for Chinese, Japanese voice for Japanese | Medium | The default for most cross-border stores |
| One voice per market | Different voices within a language (en-US vs en-GB) | Highest | High AOV, detail-sensitive categories |
Two technical notes:
- Length changes. The same sentence can differ by 30% in syllable count between languages. Leave slack in captions and shot timing — see the voiceover localization QA checklist.
- Do not translate literally. “Skin-friendly and breathable” translated word for word becomes adjective mush in English. Rewrite the hook and CTA per market. The wider process is covered in multilingual product video voiceover.
Asset specs: confirm these five before upload
| Item | Suggestion | Note |
|---|---|---|
| Aspect ratio | 1:1 or 4:5 for hero; 16:9 for mid page | Vertical occupies more screen on mobile |
| Length | 15–30s for hero | Completion drops noticeably past ~45s |
| File size | Depends on theme and plan | Check the limit your admin shows; do not guess |
| Audio | Around −16 LUFS, peaks under −1 dB | Loud tracks get compressed by platforms |
| First frame | Never a black frame | In autoplay, the first frame is the thumbnail |
One detail worth internalizing: the first frame doubles as the product media thumbnail. A black logo card wastes a free ad slot.
Nine pre-launch checks
| # | Check | Passing standard |
|---|---|---|
| 1 | Does the first 3 seconds say what it is? | Show a stranger 3 seconds; they can restate it |
| 2 | Does it work muted? | Watch with sound off and still understand it |
| 3 | Do captions match narration word for word? | Spot-check five lines |
| 4 | Are numbers and units spoken? | Verify item by item against the spec table |
| 5 | Does the pace vary? | At least three distinct paces |
| 6 | Is there exactly one CTA? | One action instruction in the whole video |
| 7 | Do locales match in length? | Within 10%, or visuals adjusted |
| 8 | Is it readable on a phone? | Captions stay within two lines on a 6-inch screen |
| 9 | Do other language versions exist? | Every locale present under the same slug |
Cost and throughput: shipping 20 videos a month
For apparel, home, and accessories, demand is many, short, and constantly revised. The pain with traditional voiceover in that pattern is not the unit price — it is turnaround and revisions.
| Item | Traditional voiceover | AI voiceover |
|---|---|---|
| Turnaround for a 30s spot | 2–5 business days | Minutes to same day |
| Changing one number | Re-record, re-sync | Edit and regenerate that segment |
| Multiple languages | Cast a talent per language | Generate every locale from one script |
| Consistency | Varies with talent condition | Stable within one voice model |
| Where it stops working | Brand films, performance-heavy long scripts | Product pages, ad variants, spec updates |
The division that holds up: brand films and emotionally demanding scripts stay with humans; product page videos and ad variants go to AI. Most mid-size stores see three to five times the monthly output from that split.
To size the cost, start with the pricing page. To lock a brand voice, use voice cloning. For volume variants, batch with TTS. For script structure — AIDA, pain points, CTAs — the conversion voiceover script guide has templates you can lift directly.
Five common mistakes
| Mistake | Consequence | Fix |
|---|---|---|
| One video for all three placements | Hero too long, bottom too shallow | Cut at least a 15s and a 45s version per product |
| Narrating every word on screen | Reading and listening fight each other | Narration states conclusions; the screen holds detail |
| One pace and pitch throughout | Sounds like a spec sheet; reads as AI | Three distinct pacing sections minimum |
| No burned-in captions | Muted autoplay delivers nothing | Export a burned-in version for hero and social |
| Literal translation across locales | Grammatically fine, unconvincing | Rewrite hook and CTA per market |
FAQ
Does a product page video need narration?
Not always. Music plus on-screen text outperforms narration in some categories (fragrance, jewelry). Whenever the video must explain “why” or “how,” narration measurably improves comprehension.
How long should the hero video be?
15–30 seconds. That slot’s job is confirming “this is what I need,” not a full introduction. Put the full story in the mid-page video.
Will AI narration sound fake?
On short, tightly contextualized product copy, synthetic narration is hard to identify. What gives it away is usually a single flat pace and unnatural pauses, not timbre. Split long sentences and leave 200–400ms between segments and most of the problem disappears.
Do we need a separate version per market?
Start with language. Same-language regions (en-US and en-GB) can usually share a video, but wording, units, and price presentation must be localized — getting that wrong damages trust more than not translating at all.
Can one video serve both the product page and ads?
Yes, with a re-cut. Ads need to be shorter, the hook has to move earlier, and the CTA must be explicit. Reusing the product page cut usually loses viewers in the first three seconds.
Does video help SEO?
Indirectly, through dwell time, conversion, and page readability. The more actionable point: your voiceover script is itself indexable product copy. Put it in a text block or collapsible section on the page and search engines have something to read.
The barrier to product video is not gear, it is script and rhythm. Get the 15-second version right first — three seconds to say what it is, ten seconds for one piece of proof, two seconds for one action — then grow it into 60. That order works far better than the reverse.


