Install Retouchable in ChatGPT Retouch your product photos from a ChatGPT chat. Install in ChatGPT
  1. Turn on Developer mode

    In ChatGPT, open Settings → Security and login and switch on Developer mode.

  2. Copy the address, open ChatGPT

    https://api.retouchable.ai/api/mcp

    Copy & open ChatGPT ↗
  3. Create the app

    On ChatGPT's Plugins page, select + and choose Create MCP App. Name it Retouchable, paste the address, set authentication to OAuth and create it.

  4. Sign in to Retouchable

    ChatGPT opens our sign-in. New to Retouchable? That starts your 14-day free trial, no card.

Then, in a chat, choose Retouchable under Developer mode and ask: “Grade the photos on my store” or “Put my product photo on white.”

Needs ChatGPT Plus, Pro, Business, Enterprise or Edu, on the web. Menus look different? OpenAI's steps

Image Load Order and Perceived Speed on Product Pages

Shoppers judge speed by what they see first, not by when the last byte lands. Here is how to order product images so the page feels fast at every tap.

|conversion optimization image optimization e-commerce

Two product pages can finish loading at the same moment and feel completely different. On one, the product appears almost at once and the rest fills in quietly. On the other, the shopper stares at a grey box while thumbnails, badges and a zoom file nobody asked for compete for the same connection. The difference is image load order, and it shapes perceived speed more than total page weight does.

This matters commercially. In Deloitte's 2020 study for Google, Milliseconds Make Millions, a 0.1-second improvement in mobile site speed was associated with an 8.4% increase in retail conversions. Most of what a shopper reads as "speed" on a product page is one question: how soon can I see the thing I came for, and how quickly does it answer when I tap?

This guide covers the parts of image sequencing that the usual performance checklists skip: what renders in the first two seconds, how to load images in response to what the shopper is about to do, how to carry the image across from the collection grid, and how to measure the speed a shopper actually feels.

Perceived speed is a sequence, not a number

Jakob Nielsen's response-time limits have held up for three decades: around 0.1 seconds feels instant, around 1 second keeps a train of thought unbroken, and around 10 seconds is where attention goes elsewhere. A product page does not need to finish inside 0.1 seconds. It needs something meaningful to happen inside each of those windows.

That is why a single load-time figure says so little. A page that shows a correctly sized placeholder in the product's own background colour at 200 ms, a recognisable product at 900 ms and a sharp hero at 1.6 s feels faster than one that shows nothing until 1.4 s and then everything at once, even though the second finished sooner.

Think of the first two seconds as a timeline you are designing:

TimeWhat the shopper should seeWhat usually goes wrong
0–300 msPage frame, title, price, and a reserved hero box filled with the product's background colourBlank white box, or no box at all so the layout jumps later
300 ms–1 sA recognisable product: a coarse or low-resolution pass of the hero, or the cached grid thumbnailThumbnails, review avatars and badges download first
1–2 sSharp hero, thumbnail strip, add-to-cart ready to useZoom-size file or slide 5 still competing for bandwidth
After first interactionThe next image the shopper is about to ask for, already fetchedEvery swipe or swatch tap starts a fresh download

The basics of getting the hero itself discovered early (never lazy-loading it, giving it high fetch priority, serving modern formats) are covered in our guide to product image LCP and Core Web Vitals. This post assumes that is done and looks at everything around it.

Image load order in the first two seconds

A mobile connection can only usefully download a few things at once. Every request that starts before the hero is a request the hero has to share bandwidth with. The job is to make sure the hero gets the pipe, and that what fills the gap before it lands looks deliberate.

Reserve the box and colour it. Set the hero's width and height (or an aspect-ratio) so the space is held from the first paint, and give the container a background colour that matches the photo's backdrop. A white box that suddenly becomes a pale grey photo reads as a flash. A box that is already the right colour reads as the image arriving.

Decide what the in-between frame is. There are three common choices, and each has a cost:

Heavier placeholders

  • Blurred low-res image (LQIP): a recognisable shape early, but an extra request or inlined bytes in the HTML
  • Blurhash-style encoded preview: no request, but needs a script to decode and paint
  • Spinners over the image area: signal waiting rather than progress

Lighter placeholders

  • Solid backdrop colour: zero bytes, works without script, near-invisible hand-off when backdrops are consistent
  • Progressive JPEG hero: a coarse full-frame pass early from the same file
  • Cached grid thumbnail: already in the browser if the shopper came from a collection page

Progressive JPEG deserves a word. It paints a rough version of the whole image first and sharpens in passes, which is very good for perceived speed. Newer formats such as WebP and AVIF usually win on bytes, and how much of them a browser paints before the download completes varies by browser and encoder. It is a real trade-off: fewer bytes and a faster finish, or an earlier recognisable frame. For a large hero on a slow connection, test both rather than assuming.

Hold back what is not part of the first view. Review photos, "you may also like" carousels, trust badges below the fold and the full-resolution zoom file should not start until the hero has. Our post on lazy loading without hurting LCP covers where the fold sits template by template.

Thumbnail strips: small files, big competition

The gallery thumbnail strip is the most common thing that quietly slows the hero. Each thumbnail is small, but a strip of eight is eight requests, often for images larger than they need to be, all starting at the same moment as the hero. On a mobile connection the hero then arrives in the time it takes to download itself plus a share of everything else.

Three fixes, in order of effort:

  • Size thumbnails for the slot. A 90-pixel thumbnail served from a 1,200-pixel file is the most common waste on product pages. Use a responsive srcset so the strip gets genuinely small files.
  • Lower their priority. Marking thumbnails fetchpriority="low" tells supporting browsers (Chromium and recent Safari) to let the hero go first. Thumbnails still load almost immediately; they just stop elbowing.
  • Start the strip after the hero. On mobile, where the strip is often below the hero or replaced by dots, it can be lazy-loaded or started on the hero's load event without anyone noticing.
Quick check

Open your product page in Chrome DevTools, throttle to a mid-range mobile profile and look at the waterfall. If the thumbnails finish before the hero starts, the order is wrong, whatever your total load time says.

The same logic applies to anything image-heavy near the top of the page: colour swatch images, badges drawn as images and logos in a promo bar. Each is a request competing with the one image the shopper is waiting for.

Load for what the shopper is about to do

Once the first view is painted, perceived speed moves to interaction. The shopper swipes to slide two, taps a colour, or pinches to zoom. If each action starts a download from nothing, the page feels slow at exactly the moment the shopper is most engaged.

The answer is to load in response to intent, not position:

InteractionSignal you can act onWhat to fetch
Gallery swipeHero finished loading; first touch on the gallerySlide 2 after the hero; slide n+1 when slide n is shown
Colour or variant switchHover or focus on a swatch (desktop); swatch row scrolled into view (mobile)That variant's hero at display size only
ZoomPointer over the hero, or the first pinchThe zoom-size file, never on page load
Quick view from a gridHover on the product cardThe display-size hero the quick view will open with

Two rules keep this from backfiring. First, only prefetch one step ahead. Fetching every variant's full gallery on hover spends the shopper's data on images they will never see and competes with the image they will. Second, prefetch the size you will display, not the largest file you have.

Variant switches are worth particular attention because they are measured. Interaction to Next Paint (INP), now a Core Web Vital, records how long the page takes to respond visually after a tap. A swatch that swaps the hero to an image already in cache paints in a frame. One that waits on the network can leave the old colour on screen while the label says the new one, which is the worst outcome: the page looks wrong, not just slow. For how swatch images themselves should be built, see our guide to variant swatch images.

Carry the image across from the collection grid

Most product page visits start from a collection page or search results, where the shopper has just seen the product's card image. That image is already in the browser's cache. If the product page's hero box shows that cached image first, at its larger size and a little soft, then swaps in the sharp hero, the product is effectively on screen the moment the page paints.

This works best when the grid image and the product page hero are the same photograph. On Shopify, the image in position 1 is the featured image: it is what collection grids and in-store search show, and usually the first slide on the product page. Keeping position 1 as the product page's opening image is what makes the hand-off seamless.

Cross-document view transitions take this further. With a few lines of CSS, supporting browsers (Chromium and recent Safari) can animate the grid card's image into the product page hero as the new page loads, so the shopper sees the product travel rather than a page reload. Browsers without support simply navigate normally, which makes it a low-risk enhancement.

One caution: a smooth hand-off makes any mismatch obvious. If the grid image is cropped tight on a white plate and the product page hero is a lifestyle shot, the transition draws attention to the jump. It is another reason the opening image should be a consistent, catalog-standard shot.

Consistent images make every placeholder trick work

Most of the techniques above depend on the images being predictable. A backdrop-colour placeholder only hides the hand-off if the photo's backdrop is actually that colour. A cached grid thumbnail only works as a stand-in if it frames the product the way the hero does. A reserved box only prevents a jump if every hero shares the same aspect ratio.

Catalogs shot over several seasons rarely meet those conditions. One batch has a warm off-white sweep, another a cool grey; some products fill the frame, others float small in the middle; aspect ratios drift between 1:1, 4:5 and whatever the photographer's camera produced. Front-end teams then compensate with per-image dominant-colour extraction, blur previews and a script to decode them, all of which add bytes and complexity to fix what is really an image-production problem.

1CSS background colour can stand in for every hero when plates match exactly
0Extra bytes for a solid-colour placeholder
1Aspect ratio to reserve when every hero is framed the same

Standardising the image set fixes this upstream. Retouchable's Studio finishes product-only shots on an exact plate colour and frames them to one standard, so a single background-colour placeholder and a single reserved aspect ratio match every product in the catalog. Pushed to Shopify in position 1, the same image becomes the grid card and the product page's first slide, which is what lets the grid-to-page hand-off look like one continuous image.

Measure the speed shoppers feel

Load-order work is easy to get wrong in ways a single score will not show. Measure it the way a shopper experiences it.

Filmstrips first. WebPageTest and Chrome DevTools' Performance panel both record a frame-by-frame filmstrip. Compare the before and after side by side and ask one question of each frame: can I tell what the product is yet? That frame is the moment the page starts to feel fast.

Speed Index for the overall feel. Speed Index, reported by Lighthouse, measures how quickly the visible area fills in rather than when the last element finishes. Better load order usually improves it even when total load time barely moves.

Element Timing for the hero itself. Adding an elementtiming attribute to the hero image lets you record exactly when it rendered, in your own analytics. It is supported in Chromium-based browsers only, so treat it as a strong sample rather than a full census.

INP for interactions. Field data from the Chrome UX Report or your own real-user monitoring will show whether swatch taps and gallery swipes respond quickly. Break it down by page type: product pages with many variants are where poor image sequencing shows up first.

Then tie it to behaviour. Faster-feeling galleries should change what shoppers do: more slides viewed, more variant switches, fewer bounces from the product page. If you test changes to the gallery itself, keep load order constant across variants, or speed will contaminate the result, a confound covered in our guide to testing product page image order.

A load-order checklist for product pages

Work through these in order. The early items cost nothing but markup; the later ones need theme or app changes.

  1. Hero discovered early and given priority. In the HTML, not injected by script, never lazy-loaded.
  2. Hero box reserved and coloured. Fixed aspect ratio, background colour matching the photo's backdrop.
  3. Thumbnails sized for their slot and set to low priority. Or started after the hero on mobile.
  4. Nothing below the first view starts before the hero. Reviews, recommendations, badges and the zoom file wait.
  5. Next slide fetched after the hero lands. One step ahead, at display size.
  6. Variant heroes fetched on intent. Hover, focus or swatch row in view, never all at once on load.
  7. Zoom file fetched on zoom intent only.
  8. Grid image and product page hero are the same featured image. So the cached copy can bridge the gap, with view transitions where supported.
  9. Measured with filmstrips, Speed Index and INP. Not just a single load-time figure.
Watch out

Theme apps for reviews, upsells and badges often inject images near the top of the product page after your theme has made its loading decisions. Re-check the waterfall after installing any app that adds visual elements, because it can undo your load order without changing a line of theme code.

Frequently Asked Questions

What is perceived speed on a product page?

Perceived speed is how fast a page feels to the shopper rather than when it technically finishes loading. It depends mostly on how soon a recognisable product appears, whether the layout stays still, and how quickly the page responds to taps like swipes and colour switches.

Should product gallery thumbnails load before the main image?

No. Thumbnails should load after, or at lower priority than, the main product image. Loading a strip of thumbnails at the same time as the hero splits the available bandwidth and delays the one image the shopper is waiting for.

What is the best placeholder for product images?

For catalogs with consistent backgrounds, a solid background colour that matches the photo's backdrop is usually best: it costs no bytes, needs no script and makes the hand-off nearly invisible. Blurred previews help when backgrounds vary, at the cost of extra bytes or a decoding script.

Should I preload every variant image so colour switches are instant?

No. Preload only the variant the shopper is likely to pick next, signalled by hover, focus or the swatch row scrolling into view, and only at the size it will be displayed. Preloading every variant wastes mobile data and competes with the images being viewed.

How do I measure whether image load order is improving my product pages?

Compare filmstrip recordings before and after, track Speed Index in Lighthouse, record hero render time with the Element Timing API in Chromium browsers, and watch Interaction to Next Paint in field data for variant switches and gallery swipes.

Give every product page the same opening frame

Retouchable finishes your catalog on one plate colour and one framing standard, so the grid image, the placeholder and the hero all line up.

Try Retouchable Free No credit card required