Optimize Product Images for Ecommerce, in Bulk, on Your Mac

Each platform publishes its own, and they disagree. We opened each page on 9 October 2026 and copied the figure.

Four things stand out. eBay publishes different caps on different sites, so a number from a forum can be right and still wrong for yours. Amazon’s own help pages sit behind the seller login, and the two Amazon-account forum posts we could open disagree on the pixel floor, so treat 1,000 px as the safe minimum and read the version in your own account; neither states a file-size cap, so we quote none. Shopify’s 2048 px is a recommendation, not a limit, and its two help pages list different formats, so test a file type before a bulk upload. And Etsy is missing on purpose: its help article refused our automated read, so we quote no Etsy figure rather than a blog’s guess. Open Requirements and Best Practices for Images in Your Etsy Shop in the Etsy Help Center yourself.

The caps catch raw exports more than compressed files: a 45-megapixel studio file is over Shopify’s 25 MP limit before its byte count matters. For any platform not listed, search its help center for the upload limits and confirm with your largest file.

Yes, for reasons other than page speed. Shopify’s documentation says its Imagery service “manages your images and videos to ensure the best possible format is displayed”, and the image_url filter detects which formats the client supports, caps width and height at 5,760 px, and never resizes an image larger than the original. An admin download returns “the original, unoptimized file”.

What your preparation still changes

• Whether the upload works at all. Anything over 20 MB or 25 MP is outside the documented limits.
• The ceiling. Shopify can only shrink. A 600 px source stays soft at every size, and for images served through image_url, a source past 5,760 px adds nothing beyond its cap. Shopify’s own advice for square product images is 2048 x 2048 px.
• The stored original. The admin hands back what you uploaded. Whatever the file carried when you uploaded it, including any GPS coordinates or camera serial, is what comes back, so strip them before upload.
• Everything that is not your storefront. Ad creative, email, a PDF line sheet and your own backups all start from the files on disk.

What it does not change: the format and widths a shopper’s browser downloads, if your theme uses image_url with srcset. Shopify picks the format for each visitor’s browser, and its performance guide leaves srcset and sizes to the theme. Before spending a week on quality sliders to fix a slow Shopify store, look at the theme and the installed apps.

Self-hosted stores are different. WordPress scales uploads over 2560 px and WooCommerce resizes to the sizes your theme asks for, all from the file you upload, so give it a clean, moderately compressed original rather than a crushed one. BigCommerce sits in between: it publishes an 8 MB cap and a short list of formats, and the pages we read do not say how it serves them.

The quality and size ranges below are our suggested starting points, not platform requirements, so test them on twenty of your own SKUs. Smol’s image defaults are quality 75, metadata stripped, and resizing to fit inside 2000 × 2000 px, so raise the maximum size for any row below that asks for more than 2000 px. Smol’s resize also has width, height and fill modes; fill crops to the box.

The one fixed fact behind 80 to 85: JPEG chroma is 4:2:0 up to quality 75, 4:2:2 at 80 to 85 and 4:4:4 from 90, so saturated edges such as a red label against a white sweep keep more color detail from 80 upward. Why AVIF for web-only images: on the 24-photo benchmark, with each tool tuned to the same SSIMULACRA2 score, Smol’s AVIF needed 22 to 26% fewer bytes than the best JPEG. That is a matched-score comparison, not a quality-65 result. Separately, at the defaults (quality 75), Smol shrank those photos 69.3% as JPEG to JPEG and 69.2% as PNG to PNG. The 65 in the AVIF row is only the quality of the benchmark’s separate PNG-to-AVIF run, so treat it as a starting point to test, not a value tuned for products. Our AVIF converter page covers the format.

Read those numbers with care. The test images are small, 768 × 512 film photographs, and we have not measured studio product photography, so your ratios will differ. Compress twenty representative SKUs, compare the byte counts, and zoom to 100% on the hardest ones.

Target file size (Neural mode) takes a size from 4 to 4096 KB and encodes toward it. It is a band, not a hard ceiling, and slow: about 28 seconds for one 768 × 512 image. Use it for the odd file that has to land near a size, then check the result against the cap; do not use it for a catalog.

JPEG for photographs, including photographs on white. Shopify’s product media page calls PNG the best file type for most product images, followed by JPEG; we pick JPEG for photographs because photographic detail generally makes smaller files as JPEG, and PNG earns its size on transparency and flat graphics. PNG only where you need transparency or have flat graphics such as size charts and badges. WebP where the platform accepts it and you want transparency in a lossy file, with one caveat on WordPress below. JPEG has no alpha channel, so a cutout saved as JPEG is a cutout flattened onto some color; flatten onto your background color in the editor, not in the compressor. Amazon main images need that flat white anyway, so a transparent cutout belongs in other slots or on other platforms.

Two traps. First, convert and compress are different operations. Smol’s convert writes JPEGs at high quality and ignores the quality setting, so for a small JPG from a PNG or HEIC use compress with the output format set to JPG. Second, format support differs by platform. iPhone shots can arrive as HEIC, which Shopify and eBay list as accepted and the Amazon posts do not. Compress those with JPG output (see the HEIC to JPG guide).

Transparent files deserve a test before a bulk run. We have not tested how Smol treats alpha on every output path, so compress one cutout, put it on a dark background and look at the edges. On WordPress and WooCommerce, test one cutout as WebP too: the WordPress 5.8 announcement said alpha formats were not yet supported for resized images, and that lossy versions are created instead. For flat graphics where no pixel may change, Smol has no lossless mode; ImageOptim does, and our head-to-head test shows where each wins. For WebP on flat product art, see converting PNG to WebP on a Mac.

Some shoots deliver TIFF or camera RAW. Shopify’s two help pages disagree on TIFF (the uploads page omits it, the product media page lists it), BigCommerce says TIF files are not supported, and eBay and Amazon accept it. Smol reads RAW (CR2, NEF, ARW, DNG) but does not develop it, so export a full-size file from your RAW tool first. Turn TIFFs into JPG before the bulk run; our TIFF to JPG guide covers sips and Preview.

Keep the master and regenerate each destination from it: a 1600 px eBay file, a 2560 px WooCommerce source, an Amazon file above 1,000 px. Each destination is a separate run, because one run produces one size, and every run writes the same name_compressed.ext by default, so move or rename each run’s outputs before the next or they collide. Compressing a compressed JPEG again stacks losses.

Set the options once, drop the top folder, review the results. The slow parts are naming and checking, not compressing.

1. 1

   Fix the file names first

   By default Smol writes a copy next to each original with a _compressed suffix. Decide your naming before the run. BigCommerce wants alphanumeric names with no spaces or special characters, and Shopify warns that endings such as “small” or “thumb” can be misread as resize requests. A hyphen or underscore may not pass BigCommerce’s alphanumeric rule, so test one file and plan to rename after the run.

2. 2

   Open Smol and set the options once

   Choose the output format, quality and maximum size from the settings table, and leave metadata stripping on. One run applies one set of settings, so make one run per destination and per format. Save each as a preset named for the destination, such as “eBay 1600” and “Shopify 2048”. Presets save the settings on that Mac, and we do not claim a way to share them between seats, so a team keeps output consistent by writing the settings down and recreating the same presets on each seat.

3. 3

   Drag in one folder per set of settings

   Whole folders and mixed file types go in one batch, but the batch uses one set of settings. A folder run at “JPG, 80” would flatten any PNG cutouts in it, so sort transparent cutouts and each destination into their own folders and run each with its own preset. For a genuinely mixed folder, set the output to the original format. Smol shows progress and savings per file as it runs and a summary when the batch finishes. On a direct compress, where compressing would make a file larger it keeps the original.

4. 4

   Spot-check the hard cases at 100%

   Open five files: a white-on-white product, one with small printed text, a saturated fabric, a fine-detail piece such as chain or mesh, and a transparent cutout. If one looks wrong, raise quality for that group and rerun only that folder.

5. 5

   Rename, then upload

   Strip or change the suffix to match your importer, check the largest file against your platform’s cap, and upload from the compressed copies. Keep the originals; they are your source for the next theme or marketplace.

Speed is honest rather than impressive: 24 small images took about 7 seconds in our test, and a sips shell loop does the same job in under one. We have not timed a 500-file folder, so run a sample first. Our Automator walkthrough covers the free route for resizing alone.

What breaks in a bulk upload besides file size

• Duplicate and unsafe names. BigCommerce requires unique, alphanumeric names without spaces. Shopify may add a unique identifier to the file name on upload.
• Embedded metadata. Phone and camera exports can carry GPS and device tags. Smol strips them by default; see removing EXIF data on a Mac.
• Color profiles. In the test behind our metadata post, Smol’s JPEG output at the default setting had no ICC tags; with metadata stripping switched off, its WebP output kept them. An untagged JPEG is normally read as sRGB, so convert wide-gamut sources to sRGB in your editor first.
• Aspect ratio and cropping. Shopify accepts ratios from 100:1 to 1:100, but says featured images only display at the same size on collection pages if the whole catalog uses a consistent ratio. Crop in the editor. Smol’s fill mode crops to a box, with no control over framing or subject.
• Accepted formats. BigCommerce rejects TIFF and PDF; the Amazon posts list no HEIC or WebP. Convert before you upload, not after the importer fails halfway through.

File names and alt text

Google’s image guidance asks for short, descriptive file names and useful alt text: my-new-black-kitten.jpg beats IMG00023.JPG, and Google discourages stuffing alt text with keywords. Name files descriptively before the run, subject to your platform’s rules above, and write alt text that describes what the shopper sees.

Small enough that the product photo does not decide your Largest Contentful Paint. Google’s threshold for a good LCP is 2.5 seconds or less at the 75th percentile, and on a product page the main photo can be the element being timed. There is no universal byte target. The levers, in order: serve the width the layout shows, use a modern format where the platform lets you, give every image width and height, and do not lazy-load the main image above the fold.

On Shopify those levers live in the theme: its performance guide recommends srcset and sizes with image_tag and warns against lazy-loading the LCP image. On a self-hosted store they live in your upload and plugin choices. We measured the width ladder and format gaps in Image Optimization for Core Web Vitals; the developer view is in image compression for web developers. If an agency maintains your theme, see Smol for web agencies.

Catalog work is not only stills. The Amazon forum post we used for image limits recommends at least six images and one video for each product, and you may also ship PDFs: spec sheets, care cards, wholesale line sheets.

Smol compresses video to MP4, MOV, WebM or MKV with H.265 (the default) or H.264, with quality labels from High (CRF 24) to Web (CRF 28) and the option to reduce resolution or remove audio. Check which codec and container your platform accepts; if you are unsure, H.264 in MP4 is the conservative pick. Our guides to compressing video on a Mac and batch video compression carry the measured results, and H.265 vs H.264 covers the codec choice.

For PDFs there are five presets, Tiny to Original. They cap the long edge of embedded images at 600, 900, 1200, 1800 and 2400 px; they are not DPI targets, and Original is not lossless. Do not plan around a percentage: PDFs exported from design tools or scanned compress well, while files from some generators barely shrink. For print line sheets read compressing a PDF for print; for a folder of supplier PDFs, batch PDF compression. Avoid Preview’s Reduce File Size filter; we measured it making PDFs bigger. No PDF preset strips metadata, so run the separate strip-metadata operation before sending a PDF to a customer.

A mix to consider is store-side delivery for shoppers plus a local pass on the files before they go anywhere. Our compressor roundup and methods comparison test the tools above side by side. Smol processes locally and uploads nothing, which matters before a launch embargo.

Solo is $29 once for one Mac. Team is $49 once for three Macs, with seats managed from an admin dashboard. Enterprise is $20 per seat, once, for 5 to 500 seats. There is no subscription and no per-file quota; details on the pricing page. The practical gain for a team is consistency: write the settings table down, create one preset per destination on each seat, and the photographer, the merchandiser and the freelancer work from the same settings. We do not document sharing presets between Macs, so the written table is the shared source.

For one-off jobs, the Finder Quick Action adds compress to the right-click menu. A watched folder can compress each new export automatically, but it has a known defect: it can make already-optimized files larger. Point it only at a folder of fresh, unoptimized exports, as described in the watched folder guide.

If someone on the team works with an AI coding agent, Smol exposes its jobs to Claude Code, Codex and Google Antigravity over MCP, so a prompt can run a compression job on a folder locally. Images that arrive on a server, from supplier portals or customer uploads, need compression there; the Smol API exists for that. Related guides: Smol for marketing teams and photographers delivering to clients.

• A Shopify store whose files are already under the limits. Shopify’s CDN already picks formats and widths. If the images are sharp and under 20 MB and 25 MP, Smol saves you storage and upload time, not a better LCP.
• Anything that must stay pixel-identical. Logos, size charts and text-heavy infographics should go through a lossless optimizer such as ImageOptim, because Smol always re-encodes.
• Automated pipelines. Supplier feeds and customer uploads are server work. An image library in the build, an image CDN, or the Smol API fits; a Mac app does not.
• Editing. Smol does not remove backgrounds, retouch or fix color. Its fill resize mode crops to a box, but with no framing or subject control, so do real cropping and aspect-ratio work in an editor first.
• Thousands of tiny thumbnails. A sips loop is faster.
• Teams on Windows or Linux. Smol is macOS only, and it does not run in CI.