For product videography, the AI tools worth paying for are a video generator for short clips of 5 to 10 seconds, a captioning and editing tool for cutting and reformatting, and an avatar presenter if you need talking explainers. Skip the all-in-one "create ads from a link" subscriptions until you have costed one finished video.
The pages that rank for this search are mostly vendor landing pages, and they all quote a price per month. What a shop needs is the price per finished video, because a plan with generous credits is expensive if three out of four clips are unusable. This tutorial works out that number for the main kinds of tool, tells you what to pay for and what to skip, and covers disclosure rules. Prices are USD list prices as checked in October 2026 and move often, so use them as a method and re-check before buying.
Start with the sum: what a finished video really costs
Take a 15-second product video made of five 3-second shots. If you generate each shot once, you need 15 seconds of footage. In practice you need more, because some shots will have warped products, odd hands or wrong motion. A fair planning ratio is three generated seconds for every second you keep. So 15 finished seconds means about 45 generated seconds.
That number multiplies against each tool's price per second. Runway's pricing page lists a Standard plan at $15 a month for 625 credits ($12 a month billed annually), which is about 2.4 cents a credit on monthly billing. It lists model costs per second of video: Gen-4.5 at 12 credits a second, Kling 4.0 at 1080p at around 23.6 credits a second (the page showed a 25 per cent Kling discount until 30 November), and Seedance 2.0 Pro at 1080p at 40 credits a second. Run the numbers for 45 generated seconds.
| Model through Runway | Credits per second | Credits for 45 seconds | Cost at about 2.4 cents a credit |
|---|---|---|---|
| Gen-4.5 | 12 | 540 | About $13 |
| Kling 4.0 (1080p) | about 23.6 | about 1,062 | About $25.50 |
| Seedance 2.0 Pro (1080p) | 40 | 1,800 | About $43 |
On a Standard plan, the 625 monthly credits cover one finished video on the cheapest model and none on the most expensive without a top-up. Move to Pro ($35 a month, 2,250 credits) and you can make about four of the cheap kind. The first lesson is that the model you pick can change the price of a video threefold, and the plan you pick decides whether that price is real or theoretical.
Google offers video generation through the Gemini API with a different pricing model. Its pricing page bills its current video model, Gemini Omni Flash, by output tokens, which the page converts to roughly 5,792 tokens for each second of 720p video, about 10 cents a second at the listed rate. By that arithmetic, 45 generated seconds comes to about $4.50, but it requires a developer-style API route, not a point-and-click app, and the page does not give resolution or audio breakdowns. Google's deprecations page also lists its earlier Veo 3.1 preview models for shutdown from 22 October 2026 at the earliest, with Gemini Omni Flash named as the replacement. Another reminder to keep your files portable.
Which tools are worth paying for, by job
Different jobs need different tools. Pay for the one that matches your most frequent video, not the one with the longest feature list.
Job 1: turn a product photo into a short moving clip
This is image-to-video: you upload your product photo and describe the motion, such as a slow orbit around a bottle or steam rising from a mug. Pay for this if you sell products where motion adds information, such as a hair dryer's airflow or a fabric's drape. A generator that you reach through a multi-model platform such as Runway lets you try several models without separate subscriptions, which is useful while you learn which one holds your product's shape best. Photoroom's Max plan, at $34.99 a month ($20.99 billed yearly) as listed in October 2026, includes a video generator that works from your product images, and it suits a shop that already uses Photoroom for stills. Verify what its video credits cover before you rely on it.
A sample prompt and the fixes it usually needs:
Start from this photo of a matte ceramic mug on a wooden table. Slow push-in, camera moves 10 per cent closer over 5 seconds. Soft morning light from the left, faint steam rising from the mug. The mug must stay exactly as in the photo: same shape, glaze colour and logo. No people, no text, no new objects.A realistic output keeps the mug mostly stable, adds believable steam, and slightly bends the handle by the last second. The fix is to shorten the clip to 3 seconds (the best part), add a still hold on the last frame from your original photo, and cut before the handle drifts. Many shops sell on the first three seconds of a clip anyway.
Job 2: presenter and explainer videos
For a talking-head explainer, such as "how to assemble this" or "how to choose a size", an avatar tool saves filming. HeyGen's pricing page lists a Free plan with three watermarked videos a month, Creator at $29 a month ($24 a month annually per its FAQ) with 600 credits and 1080p export, and Pro from $49 a month with 4K and more credits. Credits translate into video minutes in ways that depend on the avatar type, so run one real script and see what it consumes. Pay for it if you publish more than two explainers a month. Below that, a phone video of you talking to camera is quicker, cheaper and more trusted.
Job 3: edit, caption and reformat real footage
For most shops this is the best-value category. Film the product on your phone, then use an editor with AI captioning, background removal, silence trimming and automatic resizing for vertical, square and wide formats. Adobe Premiere (renamed from Premiere Pro in January 2026) and DaVinci Resolve (Studio is a $295 one-off purchase; the standard version is free) both have AI features for transcription and cleanup. For quick social clips, the editor built into a phone or a design suite is usually enough. Our guide on how videographers use AI to edit, caption and quote faster shows the workflow professionals use.
Job 4: ads made from a link
Several products promise to turn a product page URL into a finished video ad. They can be quick for a first draft, but they use stock footage and templates, so the results look like everyone else's. They are the category with the highest risk of paying for a subscription you rarely use. Test one on a free tier or a trial, and make the same video yourself with a phone and an editor. Pay only if the generated version wins on time without losing on quality.
What to skip
- Anything built on a discontinued model. OpenAI's Sora app closed on 26 April 2026 and its API on 24 September 2026. Be suspicious of any tool that still advertises it, and check which model a wrapper product really uses.
- Annual plans before a trial. Video quality changes monthly as models update; a year-long commitment locks in today's prices and features.
- Unlimited plans with fine print. Check whether "unlimited" is slow-queue, lower resolution or limited to older models.
- Tools that cannot start from your photo. A text-only generator cannot show your product, only something like it.
- Several overlapping subscriptions. One multi-model generator plus one editor beats four single-purpose tools for most shops.
A worked example: a candle range, 12 videos
Illustratively, say a small candle brand wants a 10-second clip for each of 12 products to use on product pages and social posts. Compare three routes, using the costs above and a three-to-one attempt ratio.
| Route | What it involves | Cash cost | Time |
|---|---|---|---|
| A. Phone footage plus editor | Film a flame, a pour and a hand lighting each candle; edit and caption in an existing editor | $0 to $20 for the month | About 20 minutes a video, 4 hours total |
| B. Image-to-video on Gen-4.5 | One 10-second clip per product from a still, three attempts each: 30 generated seconds at 12 credits | 360 credits per video, 4,320 for 12; about $104 at 2.4 cents a credit, which needs the Pro plan and a top-up | About 25 minutes a video, 5 hours total |
| C. Mixed | Phone footage for the hero three products; short generated loops for the rest, 5 seconds each | About $40 in credits | About 3 hours |
Route A is the cheapest and the most accurate, because the candle is the real candle. Route B brings effects that are hard to film, such as drifting smoke, but costs more than 100 dollars for a modest return. Route C is the one most small brands end up with, using AI for the shots that are awkward to film and a phone for everything that has to be true. The $/video figures are sums on listed prices, not guarantees, so replace the attempt ratio with your own.
A shot list that makes AI do less
The cheapest AI clip is the one you do not have to regenerate. A written shot list, made before you open any tool, cuts attempts sharply because each generation has one clear job. Here is a filled-in list for a 15-second video of a stainless-steel water bottle.
| Shot | Seconds | Source | Why |
|---|---|---|---|
| 1. Bottle on a desk, slow push-in | 3 | Image-to-video from your photo | Simple motion; the bottle barely changes |
| 2. Hand lifts the bottle and unscrews the cap | 3 | Phone footage | Hands and threads are where generators fail |
| 3. Water poured in, condensation beads | 3 | Generated loop | Hard to film, tolerant of small errors |
| 4. Bottle beside a laptop for scale | 3 | Real photo with a slow zoom in the editor | Scale must be true |
| 5. Logo end card with colour options | 3 | Editor | Text should never be generated |
Only one of five shots needs a generator for something truly difficult, and one more is a simple move from a still. That keeps kept generated seconds at 6 rather than 15. With three attempts each, that is 18 generated seconds, or 216 credits on Gen-4.5: about $5 at the Standard plan's rate, against $13 for an all-generated version. The remaining shots cost only your time. The wider point is that the shot list, not the tool, decides the price. It also makes the video more truthful, since the two shots where accuracy matters most, the hands and the scale, are real.
Getting usable phone footage
If part of your video is real footage, a few habits make editing and AI cleanup easier and cheaper.
- Film vertically for social, then horizontally for the product page. Reframing a vertical clip to wide loses the sides, and AI resize tools can only invent so much.
- Lock exposure and focus on the product, so brightness does not pump as your hand enters the frame.
- Use a steady surface. A cheap tripod or a stack of books beats any stabilisation tool.
- Record a clean three-second hold at the start and end of each shot. Editors use those holds for cuts and transitions.
- Record a few seconds of room tone, then let the editor's noise reduction work from it.
Caption everything. Many viewers watch with the sound off, and an AI captioning pass in your editor takes about as long as the clip itself, followed by two minutes to fix product names it misheard.
Check the product before you publish
AI video's habit of changing the product is the biggest risk for a shop. Run this check on every clip before it goes live.
- Shape. Pause on the first, middle and last frames and compare each to your photo. Handles, lids and edges drift.
- Colour. Check against the product in daylight. Video models often boost saturation.
- Text and logos. Small print and logos are the first things to warp. If they matter, overlay the real logo in the editor.
- Physics. Liquids, hands and moving parts should behave. A pour that defies gravity reads as fake.
- Claims. Does the clip imply a feature, such as waterproofing or a size, that the product does not have?
- Sound. Generated audio may include music or voices you have not licensed or approved. Replace it if unsure.
A realistic mistake, illustrated: imagine a shop that generated a clip of a leather wallet opening to reveal card slots. The model gave the wallet five slots; the product has four. Two customers wrote in to ask where the fifth slot was. Cost: two awkward emails and a reshoot. The fix is rule 5: compare every feature shown with the product's true feature list. More on trust costs is in online shop AI mistakes that hurt trust and conversions.
Disclosure and labelling
If a video looks like real footage and is not, say so. Practical steps for the platforms you are likely to use:
- Meta platforms. Meta's help page uses the label "AI content" for material it detects or that you disclose; use the disclosure control when you post realistic AI-generated video.
- YouTube. Use the "AI use" setting under Attributes when uploading realistic synthetic content.
- TikTok. Apply the AI label when posting; it cannot be removed after posting, and TikTok says it does not affect distribution.
- Your own site. A short line under the video, such as "Illustrative animation made with AI; the product is shown as photographed above", costs nothing and protects trust.
- EU customers. If you sell to people in the EU, the AI Act includes transparency duties for AI-generated content, with a grace period to 2 December 2026 for systems already on the market, as reported by law-firm briefings. Check the current position with a qualified adviser.
Never use AI to create fake customer testimonials, fabricated before-and-after results or invented people presented as real buyers. That is a legal and reputational risk, whatever the tool allows.
A first-month plan that limits the spend
- Week 1. Film three products on your phone and edit them in the free editor you have. Time the whole job. This is your baseline.
- Week 2. Take one of the three and make an AI version on a free tier or a one-month plan. Count attempts, credits and minutes.
- Week 3. Put both versions on the product page or in a post and compare views or watch time, if your platform shows them. A small sample is no proof, but a big gap is a clue.
- Week 4. Compute cost per finished video for each route, using the three-to-one ratio replaced with your real ratio. Keep a monthly plan only for the route that wins, and cancel the rest.
Since AI saves time before it saves money, write down the hours you gain and decide what they go to; why AI saves time but not money, and how to capture the gain covers that step. And if your stills are not in order, fix them first: a video made from a poor photo is a poor video, and consistent backgrounds and natural shadows sets that up. If you would like help costing a video workflow for your own range, that is what my AI implementation consultation covers, but the cost-per-finished-video sum above is enough to decide most purchases.
Questions shop owners ask about AI product video
Is Sora still an option for product video?
No. OpenAI closed the Sora app on 26 April 2026 and the Sora API on 24 September 2026. Any workflow that depended on it needs another tool. It is also a reminder that a video tool can disappear, so keep your prompts and source files outside any one vendor.
Can AI video show my real product accurately?
Only if you start from your own photo and check every frame. Generators can change shape, colour, logos and text, and they struggle with small print and moving parts. For anything a customer will compare against what arrives, use real footage and let AI handle the backgrounds, captions and variations.
How many attempts does a usable clip take?
Plan for three attempts for each clip you keep, and more for complex motion. That ratio drives your cost per finished video far more than the advertised price per second, so count it in your own trial before you commit to a plan.
Do I have to label AI-generated product videos?
Major platforms have labels or settings for realistic AI content, and some regions have transparency rules, so check each platform where you publish. Disclose when AI makes a video look like real footage of something that did not happen, and never fake customer reviews or results.
Further reads
- AI Product Photos vs a Photographer: Cost and Quality Compared — The same cost-versus-quality question for photos.
- How to Create Product Photos With AI on a Small Budget — Make the stills first, on a small budget.
- Can You Legally Use AI-Generated Images in Your Marketing? — What you can legally do with AI images in marketing.
- Taking a Local Shop Online With AI: Product Listings in a Weekend — Put products online with AI in a weekend.
- Videographer Software: Client Review, Approvals and Invoices Compared — Frame-comment review tools, proposal and invoice tools and free tiers compared for a working videographer, with dated prices and AI for feedback.
- Create a Marketing Video With AI: A Shop Owner's Short-Clip Workflow — Plan, script, film and edit a 20-second shop clip with AI help, then label it properly: platform rules, honest claims and a pre-post checklist.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Runway pricing page, HeyGen pricing page, Photoroom pricing page, Google Gemini API pricing and video documentation, the OpenAI help page on the Sora discontinuation, Meta, YouTube and TikTok AI-label help pages (checked October 2026). All prices are USD list prices and change often.