You already paid to write 200 words for every product, and not one of them has a video

Open your own store and look at what is already there. Every product has a title, a description of 80 to 200 words, three to eight photos, a price, a spec list and, on the better pages, a paragraph explaining exactly who it is for. Somebody wrote all of that. Somebody was paid to write it. And almost none of it has ever been said out loud over a moving image.

That is the strange gap in ecommerce right now. The written catalogue is finished and the video catalogue is empty. Not partially empty. On most stores carrying 40 to 200 products, the number of items with a real product video is zero, or it is the two or three bestsellers that justified a shoot once, two years ago.

The reason is never that the owner doubts video. Ask any of them and they will tell you the page converts better with one. The reason is that video has always been priced per piece, and a catalogue is not a piece, it is a multiplication. One video is an easy decision. Forty is a budget meeting.

So the useful question was never how to make a product video. It is how to make the fortieth one, and then the hundredth, at a price that survives being multiplied by the number of items you actually sell.

Narration runs at roughly 150 words per minute. That one number converts every text you own into a duration, and it is the most useful number in this entire article. Thirty seconds of finished video needs about 75 words. Sixty seconds needs about 150. Ninety seconds needs about 225.

Now hold a product description against that ruler. A common one runs 80 to 200 words. Which means a single description, alone, with nothing added, already covers a 30 to 80 second video, and at the top of that range your problem is cutting rather than padding. The script exists. It was written, reviewed, approved and paid for, and it has been on the page since launch.

Compare that with the format everybody reaches for instead. A 12 minute long form video needs around 1,800 words of continuous prose, which is a writing job in itself before a single frame is produced. That is the whole reason product video is the cheapest possible way for a store to go from zero videos to a full catalogue: the writing bill was already paid. We ran the same logic for written articles in turning written posts into finished videos.

What the description does not hand you is the order. Product copy is written to be scanned by someone who is already on the page and already interested. A video has to earn the first three seconds from someone who is not. So the rewrite is small and it is always the same: the benefit that closes the description moves to the top, the specifications collapse into one line, and the last five seconds become a single instruction about what to do next.

• 75 words of description: a 30 second video
• 150 words: a 60 second video, the length most product pages want
• 200 words: about 80 seconds, usually better cut back to 60
• 1,800 words: a 12 minute long form video, which no description covers alone

Here are the 2026 public benchmarks for hiring somebody to produce a single product video. The average price of one UGC video is US$ 212, and the common band sits between US$ 150 and US$ 300. That is for one finished piece, one product, one angle.

The tiers underneath that average matter more than the average itself. A beginner creator charges US$ 75 to US$ 300. An intermediate one charges US$ 300 to US$ 1,000. The top of the market charges US$ 600 to US$ 3,000 or more per video. And the quoted price is rarely the final one, because usage rights, meaning permission to actually run the footage in paid media, add another 30% to 50% on top of the day rate.

There is a volume discount and it is real, but it is small against the size of a catalogue. Ordering five or more videos in one package typically takes 15% to 25% off. Applied to 40 products at the market average, US$ 8,480 becomes something between US$ 6,360 and US$ 7,208, before usage rights are added back on. Nobody in ecommerce approves that in order to put a video on the long tail of the catalogue.

One clarification before the numbers go further, because these two things get mixed up constantly. This article is about the catalogue video, the one that lives on the product page, in the marketplace listing, in the shop tab and in a product Short. It is not about the paid ad creative you burn through in testing, which has its own arithmetic and its own article in what an ai avatar ad costs. Same tools, different budget line, and the catalogue one is the line that never gets approved.

Inside FalconVid the pricing unit is the credit, and one credit is US$ 0.003133. An ai avatar is billed by the second of face on screen, and the engine you pick changes the bill more than anything else you touch: Kling Avatar Std costs 32 credits per second, Kling Avatar Pro 64, HeyGen 80 and OmniHuman 1.5 costs 112. Research and script together are 307 credits, and the pipeline base is 16.

So take the standard case, a 30 second product video with the face on screen the whole way through. On Kling Avatar Std that is 30 times 32, which is 960 credits, plus 307 and 16, for a total of 1,283 credits, or US$ 4.02. The same video runs 2,243 credits on Kling Avatar Pro, 2,723 on HeyGen and 3,683 on OmniHuman 1.5, which is US$ 7.03, US$ 8.53 and US$ 11.54.

Then there is the lever that moves the number most, and almost nobody uses it. Put the face only in the 8 second hook and let the rest of the video be your own product photography with narration over it. That is 8 times 32, which is 256 credits, plus 307 and 16, for 579 credits in total, or US$ 1.81. Less than half the cost of the full face version, and on a product page it usually performs better, because the visitor came to look at the product, not at a presenter.

Now multiply, because multiplication was always the whole problem. Forty products at 1,283 credits each is 51,320 credits, which is about US$ 161 for the entire catalogue. The same 40 products bought from creators at the market average is US$ 8,480, and with usage rights added it lands between US$ 11,024 and US$ 12,720. That gap is what decides whether the long tail of your store ever gets video at all.

• Kling Avatar Std at 32 credits a second: 1,283 credits for 30 seconds, US$ 4.02
• Kling Avatar Pro at 64: 2,243 credits, US$ 7.03
• HeyGen at 80: 2,723 credits, US$ 8.53
• OmniHuman 1.5 at 112: 3,683 credits, US$ 11.54
• Face only in the 8 second hook: 579 credits, US$ 1.81

You bring the product and the text that is already written, and the pipeline handles the conversion the sections above described by hand. A researcher, a scriptwriter, a narrator, an editor and sound design run in parallel, and a finished video can be ready in up to 30 minutes.

The script comes out as prose written to be heard, not as a spec list read aloud, with the hook in front and one instruction at the end. Narration covers 63 languages with ultra realistic premium voices, so the video needs neither your voice nor your face. Karaoke style subtitles, music, sound design and 4K output are assembled by the pipeline, and the 9:16 cut is there for Shorts, Reels and TikTok without a second production.

You approve a calendar, not a script, and the videos come out on their own. The Studio lets you watch version one and adjust before anything leaves: shorten the intro, swap a photo for another one, change the music. Plans run several pipelines at once, from 2 simultaneous on Starter to 50 on Scale, which is what turns a catalogue into a week instead of a quarter.

For the 30 second avatar video costed above, the plans land like this. Starter at US$ 47 carries 15,000 credits a month, which is 11 of them. Pro at US$ 97 carries 30,000, which is 23, across up to 5 channels, and adds the dedicated server and the Senior AI Analyst. Business at US$ 297 is 95,000 credits and 74 videos across 10 channels. Agency at US$ 597 is 190,000 and 148. Scale at US$ 997 is 320,000 credits and 249, with 50 pipelines running at once. How written input becomes finished video is laid out on the text to video generator page.

There is also a free plan, 4 videos of up to 3 minutes every 30 days, with no card and the quota renewing on its own. It does not publish to a channel, you generate and download the file, and for a store that is exactly the shape you want anyway, because the file goes onto the product page, into the marketplace listing or into the ads manager rather than into a social feed.

Say the limit plainly, because pretending otherwise is how stores end up disappointed. Generative video does not hold your physical product, with your exact packaging, your exact label and your exact texture, in a real hand. If a video depends on that shot, that shot has to be filmed by somebody.

The way out is not a better prompt, it is a division of labour. The photos and clips you already own carry the object, because you already shot the product and those frames are yours. The avatar and the narration carry the words, which is the part that was expensive and slow. And the hook, the first 8 seconds, carries the sale, which is where the money on a product video actually sits.

Where a human hand is genuinely mandatory, unboxing, texture, proof of use, the correct move is to buy that piece and only that piece. A filmed insert of five seconds, cut into a video whose script, narration, editing and subtitles cost US$ 1.81, is a completely different purchase from commissioning a US$ 212 finished video for every item in the catalogue.

The other thing worth getting right is the avatar itself, because a presenter that reads like a robot damages a product page more than having no video at all. Delivery, pacing and the length of the take matter more than the engine, which is what we broke down in an ai avatar that looks real.

Most stores that ship to more than one country already translated the descriptions. That translation was an expense at the time and it is an asset now, because the script for the second and third language version of every product video is also already written and already paid for.

A project duplicates into another language paying only the difference, with its own channel and its own calendar, and narration covers 63 languages. So the Spanish version of a 30 second product video is not another full round of research and scripting, it is the narration and the render.

The question that always follows is whether three language versions of the same product video count as duplicate content, on the store and on YouTube. The short answer is no, different languages are treated as different content, and we went through the detail in the same video in three languages.

Done by hand this is exactly where every multilingual plan dies. Nobody books a creator, in three languages, for 40 products, and nobody films it themselves either. Running in parallel it is the same catalogue with another narration track, produced at the same time as the first one, in as many languages as the store actually sells in.

Day one, rank the catalogue. Sort every product by revenue over the last 90 days and draw a line. The top of the list gets the full treatment, the long tail gets the cheap version, and nothing gets skipped, because the entire point of automating this is that the long tail stops being skipped.

Day two, trim the copy. Cut each description to between 75 and 150 words, move the benefit to the front, collapse the specs into one line and end with one instruction. This is the only genuinely manual work in the week and it runs at a few minutes per product.

Day three, decide the mode per product. Bestsellers get the face on screen throughout at 1,283 credits, US$ 4.02. The long tail gets the 8 second hook with your product photography for the rest, at 579 credits, US$ 1.81. A 40 product catalogue split half and half is 37,240 credits, about US$ 117 for the whole store.

Days four and five, produce in parallel and review version one in the Studio. Days six and seven, the files go onto the product pages, into the marketplace listings and into the 9:16 cut for Shorts and Reels.

And be honest about where the ceiling really sits. Filming this by hand, a good week produces two or three finished product videos, so a 40 product catalogue is four months of work and the second language never happens at all. That ceiling belongs to the manual version. With production running in parallel, the whole catalogue is a calendar you approve once, in every language the store sells in, on as many channels as you want to run. The FalconVid home page shows the flow end to end.