How to add an audio version of your product pages and how-to pages

This page may contain affiliate links from which Blur Studio earns commissions.

Some customers read; some would rather listen while they're doing something else, driving, cooking, working with their hands. A product description or how-to page with no audio option simply loses that second group. Browser read-aloud plugins exist, but they're clunky, rely on the customer having the plugin installed, and read in a flat, machine-voiced tone that doesn't match your brand. Hiring a voice actor to record your entire catalog is accurate but expensive and slow to update whenever copy changes. If you are new to the tool, start with how to use ElevenLabs.

The accessibility case most people miss

Before you get into cost and scale, it's worth naming the reason this feature matters beyond convenience: a page with no audio alternative is genuinely harder to use for visitors with low vision, reading difficulties, or anyone relying on a screen reader that doesn't handle your page's markup well. This doesn't make an audio option a legal requirement on its own (accessibility compliance is a broader topic than one feature), but if you're already thinking about WCAG or ADA-style accessibility improvements for your site, a well-implemented audio version of your key content is a genuine, meaningful piece of that work, not just a nice extra. I'd frame the business case that way internally if you need to justify the project to anyone else, rather than purely as a convenience feature.

Start with which pages, not all of them

The mistake I'd flag before you touch any tool: don't start by generating audio for your entire catalog. Start with your highest-traffic pages, the ones that already get real visits, and prove the format works (people actually press play, it holds attention, it doesn't sound robotic) before scaling it up. This matters because the cost of this project scales with total character count across every page you cover, not with how good the idea sounds.

How to generate the audio

1. Pull your page copy as plain text, stripped of HTML, navigation text and anything that isn't meant to be read aloud (don't feed a text to speech tool your whole page source, just the actual body copy).
2. Check the character count before you generate. This is the step that actually controls your cost and matters more here than on almost any other use case, since a product catalog can run to hundreds or thousands of pages.
3. Generate the audio with ElevenLabs' Text to Speech, using one consistent voice across every page so the experience feels unified across your site.
4. Host the resulting audio file (most sites just drop it in the same storage or CDN already serving their images) and add a simple play button near the text it reads.
5. Spot-check a sample before rolling out further, since product names, model numbers and measurements are exactly the kind of text that AI narration occasionally misreads, and you want to catch that on a handful of pages, not after you've run the whole catalog.

On that last point specifically: if your catalog has a recurring pattern that trips up narration, a model-number format like "AX-4400," a unit abbreviation your industry uses a specific way, or a brand name with an unusual spelling, fix it once with a pronunciation dictionary entry rather than catching and regenerating the same mistake across dozens of individual pages. It's a five-minute setup that pays for itself the first time it stops you from re-running a batch. Part-number pronunciation is covered further in stopping a bot from reading part numbers wrong.

What this actually costs at scale

ElevenLabs' API prices by character count: $0.08 per 1,000 characters on its standard models, $0.04 per 1,000 on its faster Flash and Turbo-class models (current as of ElevenLabs' own pricing page, checked 29 October 2026). A typical product description runs 150 to 300 words, roughly 900 to 1,800 characters, so a single page costs somewhere between half a cent and a cent and a half to generate on the faster models. That sounds trivial, and it is, for one page. For a 2,000-product catalog at that average length, you're looking at roughly $70 to $150 in generation cost if you ran every single page once, before accounting for any pages you regenerate because the copy changes or the first pass needs a fix. The subscription plans cover this differently: Free gives 10,000 credits a month (enough to test on a handful of pages), Starter at $6 a month gives 30,000 credits, and Creator at $22 a month gives 121,000 credits, which is where most catalogs of a few hundred pages will land comfortably.

The number that actually bites people isn't the generation cost, it's re-generation. If product copy changes often, and in ecommerce it does, size changes, new claims, updated specs, every edit means the audio is now out of sync with the text, and keeping them in sync by hand across hundreds of pages becomes its own ongoing job.

For a smaller distributor or agency site, say 200 product or service pages rather than 2,000, the same math works out to roughly $7 to $15 for a full pass, comfortably inside the Starter or Creator plan's monthly credit allowance even before counting any free credits. The economics only get uncomfortable at genuine catalog scale, thousands of SKUs, which is exactly where the "start with your highest-traffic pages" advice above matters most. Model choice and pricing are covered in what AI text to speech costs.

If you sell into more than one language market, the same character-based pricing applies per language, so a bilingual catalog roughly doubles your character count and cost, not your page count. Budget for that upfront rather than being surprised by it after you've already committed to covering both languages.

How to know if it's actually working

Before you scale this past your test pages, put basic tracking on the play button, nothing fancy, just a count of how often it's clicked and on which pages. This is the single cheapest way to find out whether the "start with which pages" advice above actually matched your real traffic, or whether listeners are concentrated on pages you didn't expect, the ones with unusually long or technical copy, for instance. I've seen teams assume their shortest, simplest pages would get the most plays, when in practice it's the long installation guides and troubleshooting pages that get used, because that's exactly the content someone wants read to them while their hands are busy with the actual task.

Placement matters more than people expect too. A play button buried at the bottom of a long page gets missed; one placed right at the top, next to the headline, gets used. Test it in the obvious spot before assuming the feature doesn't work when the real problem is that nobody's finding the button.

My honest take for an ecommerce or distributor site

This feature genuinely helps a specific slice of customers, but I wouldn't lead your roadmap with it unless you've already seen demand for it, a support ticket asking for it, an accessibility requirement, a customer base that skews toward audio content elsewhere. Where it earns its cost fastest is on long how-to or installation-guide content, the kind someone wants read to them while their hands are busy with the actual task, more than on short product blurbs people are scanning visually anyway.

If you do roll it out, the unglamorous part, keeping audio in sync with text as copy changes, matters more to whether this stays useful than which AI voice you pick. A great voice attached to outdated copy is worse than no audio option at all, since it actively misleads a customer who's trusting it.

A smaller, easy-to-miss decision: pick one voice and lock it in before you generate your first batch of pages, not after. Switching voices partway through a catalog, because the first one you tried turned out to sound a bit off on longer pages, means re-running everything you've already done to keep the site consistent. Generate a handful of test pages with your chosen voice first, listen to them specifically on the longest, most technical page in your catalog (not your shortest, simplest one), and only commit to the full rollout once that worst-case page sounds right.

When a manual process breaks down, and what I'd build instead

For a handful of pages, generating audio by hand in ElevenLabs' dashboard is the right amount of effort. Past a certain catalog size, or the moment product copy updates regularly, doing this by hand stops scaling. What I'd build for a client at that point is a small automated pipeline: whenever a product page's copy changes, it regenerates the matching audio automatically through ElevenLabs' API and republishes it, so the audio is never more than one update cycle behind the text. That's a genuinely small, well-scoped piece of custom work, and it's the kind of thing that makes this feature sustainable rather than a one-time project that quietly goes stale. The automated version is covered in generating audio for every product in your database.