Short answer: at list prices checked on 2026-08-06, an outsourced image starts at $50, or $95 with a model, while an AI product-image subscription puts the nominal cost under $0.30 per image. Past roughly two SKUs the price gap is so wide that cost stops being the deciding variable — what should decide it is whether your product’s exact appearance can survive being re-rendered, and how much of your budget disappears into retries and subscription terms rather than into images.
This is the decision page: which path to take, and what it will actually cost. If you have already decided on the AI path and want the build, the step-by-step tooling and prompt workflow lives in AI Product Photography for Amazon Sellers.
The Compliance Floor Both Paths Must Clear
Before comparing prices, fix the spec, because a cheap image that gets suppressed costs more than an expensive one that ships.
Amazon’s main-image rules, per the Seller Central Help topic Product image requirements (G1881), require the product to sit on a pure white background (RGB 255, 255, 255) and fill at least 85% of the frame; images must be 500-10,000 pixels on the longest side, in JPEG, TIFF, PNG or non-animated GIF, with Amazon preferring the longest side above 1,000 pixels so zoom works. Main images must not carry text, logos, borders or watermarks.
Note on sourcing: the Seller Central Help hub requires a seller sign-in, so the sentence above is a paraphrase of that topic rather than a quotation, cross-read against Amazon’s own Seller Central forum guidance on 2026-08-06. Verify the current text in your own Seller Central before you brief a photographer or a prompt.
Two consequences that shape the whole decision:
- Pure white plus 85% fill is a mechanical spec. Both paths clear it easily. This is not where the risk is.
- The image must represent the actual product. This is where AI generation, as opposed to AI background replacement, gets expensive. See Gate 2.
What Each Path Actually Costs (prices checked 2026-08-06)
There are three distinct cost structures, and most comparisons collapse the middle one.
Path A — Outsourced studio shoot. You ship samples; a studio shoots, styles and retouches. Squareshot’s published prices, quoted as written (checked 2026-08-06):
| Service | Published price |
|---|---|
| Product shoot | “From $50 per image”; “$750 for a 2-hour shoot” |
| Model shoot | “From $95 per image”; “$2,950 for a 4-hour shoot”; day rate “$2,950 - $7,450” |
| Full-body model | “$175-225 for the first outfit shot”; “$85-$110 (50% off)” for each additional |
Those per-image figures are floors: complexity is rated in credits, not dollars (standard 1, advanced 1.5, complex jewelry/watches 2, hand model 3, full-body 4), with no per-category dollar price published — read the ratio, not an invoice. Non-members meet a “$300” product-shoot minimum; membership removes it, starts at “$40/month” annually, discounts 10-30%. One vendor at one price point, not a market average.
Path B — Outsourced post-production only. You shoot in-house, someone else retouches. Pixelz lists a base price of “$0.95” per image on its Professional plan (data checked 2026-08-06), with à la carte add-ons of “$0.50” to “$0.75” per image for clipping path, colour matching, invisible mannequin or advanced retouch. Base editing covers background removal, straightening, cropping, resizing and file/colour-profile handling, with a guaranteed daily capacity of 200 images and next-morning turnaround. A subscription applies on top: the page showed “$75/month” alongside “$95/month” against the annual-billing toggle when checked, so confirm which figure applies before committing.
Path C — DIY plus AI. You shoot base frames yourself and an AI tool handles background, scene and variation. Pebblely’s pricing page, checked 2026-08-06 on its monthly default, lists Lite at “US$9” for 30 images/month, Basic at “US$19” for 200, and Pro at “US$39” for 500, with annual billing advertised as “2+ months free”. Nominal per-image cost at Pro is roughly $0.08.
Photoroom is the other name sellers ask about. Its public pricing page rendered the tier structure (Pro / Max / Ultra / Enterprise, “Per month, billed yearly”, with the annual toggle defaulting on and a “Save 33%” badge) but did not render dollar amounts when checked on 2026-08-06 — the same gap our tool database recorded on 2026-07-20. A third-party cross-check dated 2026-07-20 put Pro at $12.99/month, or $7.50/month billed annually. Treat that number as unconfirmed until you see it in-app.
That is itself a decision-relevant fact: you cannot build a budget from Photoroom’s public pricing page.
The Multiplier Nobody Prices In
Sticker price per image is not spend per usable image. Two things inflate the AI side, and both show up in reviewer complaints rather than in pricing pages.
Our review corpus holds 61 verbatim Photoroom reviews (60 Trustpilot, 1 Product Hunt) dated 2025-12-29 to 2026-07-07. The sample is deliberately complaint-mined, so its rating mix (53 of 61 are 1-star) is not a representative score for the product — read it as a map of failure modes, not as a verdict. Within that sample, 44 of 61 reviews mention billing, charges, refunds, cancellation or trials, and 13 mention crashes, bugs, slowness or outright failure.
Failure mode 1 — credits burned on retries. A Pebblely reviewer (Trustpilot, 1 star, 2026-04-26) wrote: “It does not listen to instructions, and then I have to repeat again and lose credits. It changes and alter the product as it pleases.” Every retry is a paid generation. A Photoroom reviewer (Trustpilot, 1 star, 2026-06-14) reported “HORRIFIC processing time for photos, app constantly crashes, minimal support, horrible credit system”.
Failure mode 2 — the terms move under you. A Photoroom reviewer on the Pro tier (Trustpilot, 2 stars, 2026-05-21) wrote: “Yesterday I discovered that the AI credits that would renew weekly are now renewing monthly - I’ve received no notification of this change”. A Pebblely annual subscriber (Trustpilot, 1 star, 2024-03-28) wrote: “I purchased yearly plan on blackfriday deal. Soon to realise they stopped developing pebblely.ai and start promoting another AI tool as pebblely fashion.”
Failure mode 3 — trial-to-charge disputes. The largest single cluster. One reviewer (Trustpilot, 1 star, 2026-07-05) wrote: “I was charged $990 I did the free trial and a three days trial it was supposed to be $89.99 a year”. Another (Trustpilot, 1 star, 2026-06-20) wrote: “my wife went onto the app used the free trial for 3 hours and was charged £250 for the pleasure”. These are individual, unadjudicated reviewer accounts, not findings of fact — but the cluster density is the signal: pay monthly for the first cycle and set a calendar reminder before any trial converts.
So model the AI path on usable images. If you need N finished images and r is your average generations per usable image, your real unit cost is (plan price ÷ plan credits) × r. At Pebblely Pro, even a punishing r = 3 lands near $0.23 per usable image.
The Five Gates
Run these in order. The first gate that fails decides the path.
Gate 1 — Is the product itself being generated, or only its surroundings?
This is the split that matters most, and it cuts through the middle of “AI photography”. Background-replacement and scene tools take your photograph as input and change what is behind or around the product. Generative tools draw the product. The first is a retouching expense; the second is a fidelity risk. If you are on the AI path, plan on shooting clean base frames yourself — a phone, a lightbox and consistent lighting — and letting AI handle backgrounds, scenes and variants.
Gate 2 — Can your product survive being re-rendered?
Fail this gate and outsource regardless of budget. Outsource when the exact appearance is load-bearing: jewelry, watches and anything where surface, facet or stitch detail is the purchase reason; textiles where weave and drape sell the item; regulated products where labels, certification marks or ingredient panels must be legible and accurate; and any listing where a buyer comparing photo to parcel would file a not-as-described return. Note the price list above already agrees with this: those are the same categories rated 2-4x a standard image, because they are hard to shoot and harder to fake.
Gate 3 — How many finished images do you actually need?
Multiply SKUs by images per listing. Amazon listings typically carry six to nine images. This is the gate where the arithmetic gets one-sided:
| Job size | Path A (Squareshot, from $50/img) | Path C (Pebblely Pro, $39/mo) |
|---|---|---|
| 1 SKU × 6 images | $300 (hits the $300 non-member minimum) | $39 for the month, 6 of 500 credits used |
| 5 SKUs × 6 images | $1,500 | $39 for the month, 30 of 500 credits used |
| 20 SKUs × 6 images | $6,000 | $39 for the month, 120 of 500 credits used |
At one SKU the comparison is $300 against $39 plus your own shooting time — close enough that convenience wins. At twenty SKUs it is $6,000 against $39. Cost has stopped being an argument; Gate 2 is doing all the real work.
Gate 4 — What is actually blocking your launch?
Outsourcing has a physical critical path that no budget removes: pack samples, ship them, wait for a slot, shoot, retouch, revise. Shipping and scheduling — not shutter time — dominate that timeline, and a revision round restarts the queue. The AI path front-loads the work into your own hands and your own retries. If a launch date is fixed and samples are not yet at a studio, that constraint outranks price.
Gate 5 — What is your billing exposure over twelve months?
Path A is transactional: you pay per job, and an unused month costs nothing. Path C is a subscription with the failure modes above. Reduce exposure mechanically — pay monthly until a tool has earned an annual commitment, use a card you can dispute, screenshot the plan terms you signed up under, and cancel through the platform that took the payment (App Store subscriptions cancel in the App Store, not in the app).
Where the Two Paths Cross
For a seller with more than two SKUs and no fidelity-critical categories, the AI path wins on cost by an order of magnitude, and the decision should be made on Gate 2, not on the invoice. For a single hero SKU in a fidelity-critical category, one outsourced shoot at $300-$600 is the cheaper decision once you price a suppressed listing or a returns spike.
The hybrid is what most sellers land on and it is worth naming: shoot one real, correct set of base frames — outsourced or in-house — then use AI for every background, scene and seasonal variant afterwards. It buys fidelity once and amortises it across every subsequent image. Path B (in-house shoot, outsourced retouch at roughly $0.95-$1.70 per image all-in) is the same logic with human hands on the retouch.
Before any of this, look at what already ranks in your category. Pulling the full image set from the current top listings tells you which shot types the category actually rewards, and how many images you genuinely need — a browser tool like ASINCrate exports every image from a listing in bulk for exactly that purpose. Budget the shot list you need, not the one a pricing page assumes.
Common Mistakes
- Comparing sticker prices instead of cost per usable image. Apply your retry multiplier before the comparison means anything.
- Buying an annual AI plan in month one. The corpus above is full of annual subscribers whose terms changed afterwards.
- Letting a generative tool redraw a product whose details are the product. Gate 2 exists because this failure surfaces as returns, not as a rejected upload.
- Forgetting the minimum order. A $50/image list price does not mean a $50 invoice; $300 minimums are normal.
- Ignoring sample logistics in the timeline. Shipping and scheduling, not shooting, set the outsourced calendar.
- Skipping the compliance check on AI output. Verify pure white, 85% fill and longest side above 1,000 pixels on every file before upload.
Quick Checklist
- Count finished images needed: SKUs × images per listing.
- Answer Gate 2 honestly — is exact appearance load-bearing for this product?
- Price Path A against your real shot list, including the vendor minimum.
- Price Path C on usable images, not nominal credits.
- Check the launch date against sample shipping time.
- Choose monthly billing for the first cycle of any AI subscription.
- Validate every final file against Amazon’s main-image spec before upload.
- If you chose AI, follow the build steps in the workflow guide, and keep image metadata clean on export.
Frequently Asked Questions
How much does Amazon product photography cost in 2026?
At list prices checked 2026-08-06, outsourced studio images start at $50 for a product shot and $95 for a model shot, typically over a $300 minimum order. Outsourced retouching of your own photos runs about $0.95 per image plus add-ons and a monthly subscription. AI product-image subscriptions run $9-$39 per month for 30-500 images, putting nominal cost under $0.30 per image.
Does Amazon allow AI-generated product images?
Amazon’s image rules govern the result, not the tool: the main image must show the product for sale on a pure white background, filling at least 85% of the frame, with no added text or graphics. The compliance risk with generative AI is not the technique but accuracy — an image that misrepresents the actual product creates a listing-accuracy problem regardless of how it was produced.
Is AI product photography cheaper than hiring a photographer?
On unit cost, dramatically — by roughly two orders of magnitude at volume. But the AI path assumes you can produce clean base frames yourself and that your product tolerates re-rendering. For jewelry, watches, textiles and regulated products, one correct outsourced shoot is usually the cheaper decision once returns and suppression risk are priced in.
What is the hidden cost of AI image subscriptions?
Retries and subscription terms. Failed generations consume paid credits, and reviewers in our 61-review Photoroom sample (dated 2025-12-29 to 2026-07-07) reported credit renewal periods changing without notice and disputed trial-to-paid charges — 44 of the 61 reviews touch billing in some form. Budget on usable images and pay monthly until a tool has proven itself.
The Verdict
Cost is the question sellers ask and the wrong axis to decide on. Below two SKUs, either path is affordable and convenience decides. Above two SKUs, AI is so much cheaper that the only question left is Gate 2 — whether the product’s exact appearance is the thing being sold. Answer that first; the invoice follows from it.
For a broader look at how photography services were positioned before generative tools existed, the older service-provider roundup at amzfinder.com covers the traditional vendor landscape.
Prices in this guide were checked on 2026-08-06 against each vendor’s public pricing page and change frequently, including by region and billing toggle. Verify current pricing before committing.