If you have ever tried to right-click an Amazon product image and got back a 500-pixel JPG with a name like 71xKf-mdsTL._AC_SL500_.jpg, you have already met the problem. Amazon’s image pipeline is built to deliver pre-sized, pre-cached derivatives — never the original — and the carousel hides the variant set inside a JSON blob most people never look at.

This guide is the technical reference I wish existed when I first tried to build an Amazon competitor research workflow. We will decode the image CDN URL grammar, walk through the four real download methods (with the trade-offs nobody mentions), then translate the dump into something useful: a CTR-focused image audit framework you can run on the top 20 listings in your category.

How Amazon’s Image Pipeline Actually Works

Amazon does not store one file per image. Every uploaded product photo is processed once on ingest, then served on demand through an imaging service that crops, resizes, recompresses, and watermarks on the fly. The URL itself is the API call.

Three things are useful to know before you try to download anything:

1. The original is always reachable if you strip the size flags from the URL. The CDN keeps the source asset; the URL just tells the renderer what derivative to produce.
2. There is no single image gallery in the DOM. The visible carousel only renders the currently selected variant. The full set lives in a JavaScript object called colorImages (or initial.colorImages.initial on some templates) — a hash where each key is a variant ASIN and the value is an array of image objects with hiRes, large, thumb, and mainUrl fields.
3. A+ Content uses a separate host and a separate DOM region. It is not part of colorImages. You have to scan #aplus_feature_div (or the equivalent aplus3p_feature_div for vendor brands) and pull the <img> tags from there.

Most “amazon image download” guides you will find online stop at the carousel. Half the assets on the page never get pulled. That is the gap between an extension that gets you 5 images and one that gets you 38.

The Amazon Image URL Grammar

The image URL is the most useful thing you can learn to read. It looks chaotic; it is actually a small DSL. Take a real example:

https://m.media-amazon.com/images/I/71xKfQk-mdsL._AC_SL1500_.jpg

There are four meaningful parts:

Segment	Meaning
m.media-amazon.com/images/I/	The current global image host. (Older listings still use images-na.ssl-images-amazon.com — both work.)
71xKfQk-mdsL	The asset ID. This is the only piece tied to the specific image.
._AC_SL1500_	The processing block. Each flag is dot-separated.
.jpg	Output format. Can be replaced with .png or .webp on most assets.

To get the original, you keep the asset ID and drop the processing block:

https://m.media-amazon.com/images/I/71xKfQk-mdsL.jpg

That URL returns the source upload, which for current Amazon listings is usually 1500–2560 px on the longest side, sometimes 5000+ px for A+ Premium modules and 360-spin frames.

Every Processing Flag, Decoded

Below is the working flag list. You rarely need all of them, but understanding them makes the next time Amazon changes their CDN behavior a non-event.

Flag	Effect
_AC_	“Amazon Common” — the standard rendering context. Often present but not strictly required.
_SL1500_	Scale Longest side to 1500 px. Replace number to control size.
_SX466_	Scale to width 466 px (height auto).
_SY679_	Scale to height 679 px (width auto).
_SR300,300_	Scale and Resize to a 300×300 box, padding if needed.
_CR0,0,500,500_	Crop Rectangle: x, y, width, height in pixels of the source.
_QL80_	Quality Level 80 (JPEG quality). Default is around 70.
_FMjpg_	Force output format to JPEG (also _FMpng_, _FMwebp_).
_PIfont,TopLeft,...	Programmatic Image overlay — used for “Best Seller” or “Limited Time” badges.
_US40_	Upper-left Sprite — internal use, mostly safe to remove.
_RI_	Render Image flag, usually a no-op for derivatives.

The order matters. Flags are read left to right and chained, so ._SX300_AA300_QL80_ resizes to 300 wide, then constrains to a 300-square frame, then re-encodes at JPEG quality 80.

The “Trick” That Half the Internet Recommends

You will see blog posts that say “just change _SL500_ to _SL1500_ to get the high-res version”. That works, but it is the long way around. The faster move is to remove the entire processing block. The original is always there — you do not have to guess the maximum width.

Two practical caveats:

• Some very old listings store source images smaller than the modern requirement. If you strip flags and get a 1000 px image back, that is the original.
• Variant swatches (_SS40_) are intentionally small thumbnails. Stripping the flag returns a 200–300 px image, not a hi-res one. Swatches are a different asset class.

Where Each Image Type Lives

Amazon’s filename suffixes are a cataloging system, not a coincidence. When you pull an organized download you will see codes like MAIN, PT01, SWATCH — they map to the slot the image fills.

Code	Slot	What sellers use it for
MAIN	The hero image — only one per listing. Pure white background, product fills ≥85% of the frame, no text or badges allowed.	The single biggest CTR lever. The frame ratio rule is enforced.
PT01–PT08	The 8 secondary slots. No background restriction.	Lifestyle, infographics, dimension diagrams, ingredient detail, before/after.
SWATCH	Variant-picker thumbnails (color, pattern, finish).	The little chips next to the variant selector.
APLUS_*	A+ Content modules below the bullet points.	Brand story, comparison charts, banner panels.
APLUS_PREMIUM_*	A+ Premium (Brand Registry only): 1464 px wide hero, video carousel, hotspot images.	Higher-fidelity brand storytelling. Very few categories use it well — often a low-hanging differentiation move.
IN01–IN08	“In-context” alt images, surfaced in search-result side previews on some marketplaces.	Lifestyle context Amazon may pick to show in search.
REVIEW_*	Customer-uploaded review photos. Owned by the reviewer.	Real-world product photos, packaging shots, defect documentation.

When you study a competitor, the distribution across these slots tells you their strategy. A listing that uses MAIN + PT01–PT07 + APLUS_PREMIUM is investing real money in conversion. A listing with MAIN + 3 PT slots and no A+ is leaving CTR on the table.

Method 1: Chrome Extension (the Practical Default)

A purpose-built extension parses the colorImages JSON, walks the A+ DOM region, builds the bare URL for each asset, and packages everything into a ZIP with per-variant folders. That is the entire job. The reason this is the default for most sellers is not capability — every method below can technically retrieve a hi-res image — it is throughput. A 12-variant listing with full A+ Content and 6 review videos is a 90+ asset operation. Doing that one URL at a time is not a workflow.

How to Use ASINCrate

1. Install ASINCrate from the Chrome Web Store. Also works on Edge, Brave, and Arc.
2. Open any Amazon product page on the 15+ supported marketplaces (US, UK, DE, JP, IN, FR, IT, ES, CA, AU, MX, BR, NL, SE, AE, SG, TR, PL).
3. Click Download ZIP. The extension grabs MAIN, all PT*, every variant’s full image set, all SWATCH thumbnails, every A+ Content image, and customer review media. Files land in folders named by variant; filenames include the ASIN and the slot code.
4. Optionally export a CSV with columns for ASIN, variant, slot, hi-res URL, dimensions. This is the input format you want if you are running an audit across 20+ listings.

What you do not get with right-click: per-variant organization, slot codes embedded in the filename, A+ Content, review media, and a manifest you can pivot in a spreadsheet.

Other Extensions, Honestly Compared

• AMZ Downloader — Free image export; videos and CSV behind a $6.99/mo tier. Does not parse colorImages for hidden variants on every template.
• SellerSprite AmzSave — Requires a SellerSprite account login. Image-only, no A+ Content separation.
• EcomStal — 100% free, broad marketplace coverage, but no CSV export and no variant-folder structure.
• ASINCrate — Variant-folder ZIP, CSV manifest, A+ Content + review media, HLS video download, no account for single downloads.

If the only thing you ever need is “pull the main image once”, any of these (or right-click on the unsized URL) is fine. If you are doing weekly category audits, the difference is hours per session.

Method 2: Right-Click Save (Why It Almost Never Works for Sellers)

You can right-click any visible image and pick “Save image as”. You will get the rendered derivative, not the original — usually _SL500_ or _AC_SL679_, depending on viewport. The filename will look like 41ABC123._SL500_.jpg, with no ASIN and no slot context.

Even with the URL trick (open in new tab, strip the size flag, save), you only get the currently displayed variant. The other 11 variants never load into the DOM unless you click into them. This is the single biggest reason right-click fails for serious research: the page hides ~80% of the assets.

Use it for: pulling one image as a quick reference. Skip it for: anything systematic.

Method 3: Web Scraping (When You Are Doing 1000+ ASINs)

For data-team workflows pulling thousands of ASINs into a database, dedicated scraping infrastructure makes sense:

• Crawlbase / ScraperAPI / Bright Data — Provide rotating residential or datacenter proxies and CAPTCHA solving. You write the parser; they handle the request layer. Expect $0.001–$0.005 per request at scale, plus engineering time to maintain selectors as Amazon’s HTML changes.
• Keepa / Helium 10 API — Higher-level product data APIs. Image URLs are part of the response. Cheaper per ASIN but lower fidelity (typically the main image only, occasionally up to 7 secondary).
• ImportFromWeb — Google Sheets formula that pulls Amazon HTML into a cell. Gets you URLs to copy-paste; not actual file downloads.

These are infrastructure choices for a data engineering team, not a workflow for an individual seller. The trade-off is real: you get scale, you give up A+ Content fidelity, and you take on maintenance burden when Amazon updates the page template.

Method 4: The DevTools URL Trick

The “free” method for technical users:

1. Open the listing, hit F12, go to the Network tab, filter on media-amazon.
2. Click through every variant and view all 7 product images. Each request is logged.
3. For any logged image URL, copy it, strip the processing block (everything between the asset ID and the file extension), and paste the cleaned URL in a new tab. Save.
4. To get A+ Content, scroll the page to trigger lazy-loaded images, then re-filter Network on aplus.

This works. It is also a 30-minute exercise per listing. Worth knowing the technique because it tells you what an extension is actually doing under the hood; not worth doing by hand for any volume above 1–2 ASINs.

Method Comparison

What to Actually Do With the Images

Most posts about amazon image download stop here. The interesting work starts after you have the files. Here is the audit workflow that the better Amazon agencies run on a new client account.

Step 1: Build a Category Image Library (one afternoon)

Pick the top 20 listings in your category. Pull all images for all 20 with a CSV manifest. You should end up with roughly 800–1500 image files. Sort the CSV by slot, then by competitor.

Step 2: Audit the Main Image

Open the 20 main images side by side in a contact-sheet view. Look at five things:

• Frame fill. Amazon’s rule is ≥85%. A surprising number of competitors are at 60–70%. Yours should be 85–95%, no margins.
• Angle. Most categories converge on one angle (front-elevated 15° for kitchen tools, true-front for apparel, three-quarter for electronics). The outliers are either category leaders who can ignore the convention or laggards who do not know it.
• Color and contrast. Sort by dominant color. If 18 of 20 listings are visually identical, the differentiation play is to break the pattern.
• Packaging vs. product-only. This is a real split. Subscription/consumable categories (supplements, food) often show packaging; durable goods rarely do. Match the dominant pattern unless you have a specific reason to break it.
• Implicit scale cues. Hand-held, tabletop, on-shelf — does the main image telegraph size? In categories with frequent “smaller than expected” returns, this matters.

Step 3: Reverse-Engineer the PT Slot Strategy

Build a 8×20 grid: PT01–PT08 across the top, the 20 ASINs down the side. Mark each cell with the image type (lifestyle / infographic / dimension diagram / ingredient / before-after / size comparison / packaging / video thumbnail). Patterns surface in 30 seconds:

• If 18 of 20 listings have an infographic at PT01, that is the convention. Skip it and you are missing the “what is this” moment.
• If only 3 of 20 use a dimension diagram and your category has frequent sizing complaints in reviews, that is your unfair advantage.
• If a PT slot is consistently empty across the category, fill it. Amazon will surface it in search; competitors are leaving the slot to you.

Step 4: Score the A+ Content

A+ Content is heavily under-used. In most categories, fewer than half of the top 20 listings have a meaningful A+ module — and of those, most use the default templates with stock-style imagery. A custom-designed A+ Premium module (only available with Brand Registry) is one of the cheapest CTR wins still available because so few sellers do it well.

The audit question: of the top 5 listings, how many have branded A+ Content (custom photography, comparison chart vs named competitors, brand story, founder photo)? If the answer is zero or one, that is the gap.

Step 5: Mine the Review Photos

Customer review photos are the single best source of honest product feedback you will ever read. Pull the review media for the top 5 competitors. Look for:

• Defect patterns. If 30% of review photos show the same broken plastic clip, that is your QC angle.
• Real-world scale. Customer photos almost always reveal that the product looks smaller in real life than in the marketing photos. If your competitors are over-styled, an honest scale shot is differentiation.
• Use cases the seller did not anticipate. Reviewers often photograph creative uses. Add those as PT slots in your own listing.

Marketplace Differences That Catch People Out

The image system is the same across marketplaces, but the conventions and the rules are not.

• JP rewards higher fidelity. Top JP listings consistently use 2560+ px main images and aggressive infographic density. Western copies of JP brand listings often look low-effort by comparison.
• DE has stricter compliance review on text-on-image. Claims like “Bestseller” or “100% Natural” embedded in PT slots get flagged more often than in US.
• IN allows currency and price overlays in PT images more freely than US, which is otherwise discouraged.
• UK / EU has the same TOS as US on watermarks (forbidden on MAIN) but enforces it less consistently.
• Brand Registry unlocks A+ Premium, video in the carousel, and Sponsored Brands video — a meaningful gap between Brand-Registered and unregistered listings in any marketplace.

Image Compliance Pitfalls That Will Kill Your Listing

Amazon’s main-image policy is mechanically enforced (image classifier + manual review on appeal). The frequent-violation list:

1. Main image with text, badge, watermark, or border. Auto-rejected. Includes “Made in USA” badges, brand logos floating in the white space, “Best Seller” overlays.
2. Main image with a person’s face when the product is not apparel/cosmetics. Often rejected.
3. Main image with packaging when the product is not normally sold in the box (i.e., the box is not the unit). Inconsistently enforced, but flagged in audits.
4. Main image where the product fills less than 85% of the frame. Auto-detected.
5. Main image not on RGB (255,255,255) pure white. Off-white or shadowed backgrounds get caught.
6. Lifestyle PT image with hands or scenes that imply medical, weight-loss, or “before/after” claims in regulated categories (supplements, beauty). Subject to suppression.

Run your own listings through the same rule check before shipping a redesign. The number of suppressed listings caused by an over-eager designer adding “Top Rated” to the main image is genuinely large.

The Legal Boundary

Downloading is the easy part. The boundary is what you do with the file:

• Reference and analysis — fine. Same as walking into a competitor’s store with a notebook.
• Direct reuse in your own listing — copyright infringement. Amazon will action a complaint within 48 hours and can suspend the listing.
• Substantially similar derivative work — the same composition, same prop, same angle, same color treatment is enough to trigger a takedown even without pixel-identical copy.
• Customer review photos — owned by the reviewer, not by Amazon and not by the seller. Internal analysis only.

A safe rule: anything that would feel awkward to put in a slide deck shown to the competitor’s brand manager is probably the wrong side of the line.

Frequently Asked Questions

What resolution does Amazon actually store product images at?

Amazon keeps the original upload, typically 1500–2560 px on the longest side for current listings — and up to 5000+ px for newer A+ Premium and 360-spin assets. The page only renders compressed derivatives (usually 500–679 px), so the in-page DOM never exposes the original. You either request the unsized URL directly or use a tool that builds it for you.

What do filenames like 71xKf...AC_SL1500_.jpg actually mean?

The filename has two parts: the asset ID (the random characters before the first underscore) and a series of size and processing flags. _SL1500_ means scale longest side to 1500 px; _AC_ is the standard rendering context; _UX466_ means scale to width 466. Removing every flag and the trailing block returns the original.

Can I download every variant image without clicking through them?

Yes. Amazon stores the full variant image set inside a JSON object called colorImages (or initial.colorImages.initial) embedded in the HTML. ASINCrate parses this object directly, so it captures all variant images in a single download instead of forcing you to navigate variant by variant.

How do I download A+ Content and A+ Premium module images?

A+ Content images are served from a different host (typically m.media-amazon.com/images/S/aplus-media or aplus-media-library-service). They are not part of the main image carousel JSON. ASINCrate scans the A+ DOM separately and includes those assets in the same ZIP, labeled APLUS_*.

Is bulk downloading Amazon product images against the TOS?

Amazon’s TOS forbids automated scraping that hits their servers at scale. Loading a single product page in your own browser and saving the images it already delivered is a different category — it is the same network traffic the page already produced. The legal risk is not the act of saving; it is what you do with the file. Using a competitor’s image inside your own listing is a copyright infringement Amazon will action.

Conclusion

The image pipeline is built for delivery, not for download — which is why a 5-minute right-click attempt feels harder than it should. Once you understand that the URL is an API call, that the variants live in a JSON object, and that A+ Content has its own host, the whole thing becomes mechanical.

For one-off curiosity, the URL strip trick is enough. For weekly competitive audits across 20+ listings — the actual seller use case — a Chrome extension that parses colorImages, walks the A+ DOM, and writes a CSV manifest is the only workflow that scales.

Need to grab the videos too? See our companion guide on downloading Amazon product videos, including how the HLS/m3u8 streams and signed token URLs actually work.