How Long Does Virtual Try-On Take on Shopify?
How long virtual try-on takes on Shopify: generation wait vs page load. Antla publishes no median. Dated vendor seconds and a trial timing method. 24 Oct 2026.
Photo try-on is not instant. After the shopper uploads a still on the product page, they wait for a generated picture of that SKU on their body. Antla does not publish a median generation time. Seconds on other vendor pages are listing copy. Time the wait on your catalog during the 7-day trial.
See a clothing preview at Antla.io/demo. That page is a booking link, not a stopwatch.
The search query sounds like a single number. It is three different clocks, and most landing pages answer the wrong one. I would not put “results in X seconds” on a product page until I had timed this catalog, on a phone, on a mid-range connection. Vendor marketing is not that measurement.
Three clocks get mixed into one question
Shoppers asking “how long does virtual try-on take” usually mean the pause after they tap Try on. Merchants asking the same sentence often mean “will this tank Lighthouse.” Those are not the same product.
Clock one is first load. Hero photo, price, add to cart. Google’s Core Web Vitals score that at the 75th percentile, split by mobile and desktop: Largest Contentful Paint within 2.5 seconds, Interaction to Next Paint of 200 milliseconds or less, Cumulative Layout Shift of 0.1 or less. Generation does not start until someone uploads. Shoppers who never tap Try on never pay clock two. The storefront version of this objection is already written: does virtual try-on slow down a Shopify store. This page stays on the wait after tap.
Built for Shopify is a clock-one filter, not a generation SLA. Shopify’s Built for Shopify requirements say a storefront app “must not reduce the storefront Lighthouse performance score by more than ten points.” That is a badge rule. It is not a published Antla render time.
Clock two is the upload. Camera permission, camera-roll picker, a bad crop, a retry. On a train, that can outlast the model. The shopper flow is in upload a photo to try on clothes. Time it separately from generation. If the photo is already in the session, the second SKU should not make her pick a file again. That is upload time saved, not a published generation median.
Clock three is generation. Two stills in, one still out: her photo and your product image. Shopify’s 2026 virtual fitting room guide stacks live AR, AI, and 3D under one retail phrase. Live AR keeps the device busy for as long as the camera is open. Photo try-on does the heavy work after upload and returns a still. Do not grade a clothing still against a glasses overlay that never leaves the camera.
A number that does not name which clock is a slogan.
Listing seconds are not a lab
I opened the vendor pages below on 24 October 2026. I did not time any of them. Treat the cells as quoted marketing, then run the same five SKUs on a duplicate theme.
| Source opened 24 Oct 2026 | What they print about wait | What that is |
|---|---|---|
| Antla listing, pricing, Pro AI, FAQ, docs | No median generation time. Unlimited Fashion names “Priority Image Gen.” Headless docs say poll about every two seconds; etaSeconds is an estimate, not a guarantee. | Public product copy. Not a lab. |
| TryOnCloud App Store listing | Bullet: shoppers “see clothes on themselves in 7 seconds” | Listing copy |
| Photta Business FAQ | ”within ~10 seconds.” Same page also says “See it work in 12 seconds.” | Widget copy. Photta’s separate catalog-image API is a different clock. |
| Looksy product page | ”Try-on time” listed as 20 seconds | Product-page copy |
| GenLook outerwear page | ”8.2s median time per try-on” | Marketing median |
| GenLook changelog, 20 Oct 2026 | August 2026 median 12.5 s. 6 to 19 Oct 2026 median 5.1 s, measured photo-send to result | Same vendor, different window |
The useful fact in that table is not “7 beats 20.” It is that one vendor’s own published median moved inside a quarter, while an older category page still showed 8.2 seconds on the day I opened both. If you need a category number for a board slide, you do not have one. You have dated screenshots.
Photta’s shopper widget and Photta’s on-model catalog API are easy to smash together. The Business FAQ is the ~10 second line. Photta’s apparel API docs talk in minutes. Quote the surface you are actually installing.
Looksy’s page also sells a size recommendation in the same breath as the wait. Antla has no size chart or size recommendation of its own. It works alongside any Shopify size app. A 20-second claim that includes a size pick is not the same job as a clothing still.
I am not repeating vendor conversion or return percentages from those pages. They are not a stopwatch.
What Antla actually publishes about the wait
Opened 24 October 2026, none of these print a median in seconds: the Shopify App Store listing, the pricing page, antla.io/faq, virtual try-on, or Antla Pro AI. Anyone quoting “10 to 30 seconds” as an Antla number made it up. The how-it-works rewrite already says that: how AI clothing try-on works.
What is on the public pages:
- Fast vs Pro. The Pro AI page says merchants switch between Fast and Pro from the dashboard. Fast is the quicker path. Pro is the higher-fidelity path. No seconds for either.
- Priority Image Gen is a named Unlimited Fashion feature on pricing and on the App Store. It is a plan line, not a published millisecond.
- Headless polling. Custom UI workflow says
etaSecondsis an estimate, not a service-level guarantee. Poll about every two seconds. A suggested UI deadline is three minutes, “chosen by your app; it is not an Antla guarantee.” Two seconds is how often a custom UI asks “done yet.” Three minutes is a client timeout you pick. Neither is a median. - Job states. Poll a generation documents
PROCESSING(keep waiting), a temporary preview you may show while storage is still pending, then a final URL. A hung spinner with no deadline is a product bug you shipped, not a model SLA.
The 7-day trial is the measurement window. Trend is $19.99/month for 100 try-ons, then $0.16. Runway is $49.99/month for 500, then $0.12. Unlimited Fashion is $199.99/month for 2,000, then $0.10. There is no free plan.
Antla’s privacy policy, opened 24 October 2026, states: Images are deleted and not accessible within 72 hours. Link: antla.io/privacy-policy. The wait on screen is not a 72-hour hold. The 72-hour line is deletion. Do not tell shoppers the file is never stored, and do not invent a training claim.
Time it on a duplicate theme, not on a homepage GIF
I would not accept a sales-deck second. I would time the path the shopper actually walks.
- Duplicate the live theme. Put try-on on Default product only. Confirm the script is absent from collection and cart. Clock one lives there.
- Pick five SKUs you will actually sell this month: a simple tee, a print, a midi dress, a dark wash, a jacket. Feed the app the front packshot you will ship, not a back crop.
- Use a mid-range Android on a throttled 4G profile, not the office fiber and not the demo laptop.
- Start the timer when the upload confirms, not when the page opened, and not when she tapped Try on. Stop when a still is on screen. Write the three clocks in three columns.
- Run Fast and Pro on the same five SKUs with the same shopper photos. You are scoring wait and whether the print still looks like your print. A faster still that invents a logo is a fail.
- Generate a second SKU without picking a new photo. If she is sent back to the camera roll, that is an upload defect. If generation itself is slower on SKU two, write that down. Do not average it into a fake median.
- Repeat the worst SKU three times. One lucky 4-second render is not an SLA.
Record failures. A timeout, a blank still, or a retry is part of perceived wait. Vendors who publish only completed-try-on medians have already dropped the jobs that bounced.
Jakob Nielsen’s response-time limits (1993, still the web-app rule Nielsen restated in 2014) are the reason this QA is worth an afternoon. 0.1 seconds feels instant. 1 second keeps the flow of thought. 10 seconds is about the limit for keeping attention on the dialogue. After that, people want to do something else while they wait, and they need a way back. Photo try-on is not a 0.1-second control. Design the wait as a wait. Do not pretend it is a toggle.
I would not copy a “percent done” bar that lies. Nielsen’s own rule is that a fake progress indicator is worse than a spinner that at least admits the work is unbounded. Antla’s public docs do not publish a percent-complete. Do not invent one in theme copy.
Keep add to cart usable while the still cooks
The failure mode I see is a modal that owns the whole PDP until the model returns. She cannot read the fabric line. She cannot open the size chart. She cannot add the size she already knew. Then the still is late, she is annoyed, and the only button left is close.
Leave add to cart, variant, and the size app tappable. The preview is extra evidence, not a gate. If generation fails, she should still be on a working product page.
Put one honest line next to the spinner: you are making a picture of this item on her photo; it is not a size; she can keep shopping. Do not promise a second count you have not timed on this catalog.
After 10 seconds of attention, Nielsen’s point is reorientation. When the still lands, put the next action next to it: add this variant, try another color, close. A result that appears over a page she has already scrolled away from is a miss even if the model was “fast.” Placement after the wait is in information beside the virtual try-on preview.
Install order, including which SKUs to start with, is how to add virtual try-on to Shopify.
If you came from a comparison page that already said “listing copy, not a lab,” this is the operator version of that sentence: Looksy vs Antla, Photta vs Antla, TryOnCloud vs Antla.
Merchant questions
How long does virtual try-on take?
On a Shopify clothing product page, photo try-on takes as long as the upload plus the generation of a still. Antla does not publish a median. Other apps print 7, ~10, or 20 seconds on pages we opened 24 October 2026. Those are listing claims. Time the wait on your catalog.
Is generation time the same as store speed?
No. First load is Core Web Vitals. Generation starts after the shopper uploads. A vendor quoting a render time is not answering Lighthouse.
How long does Antla take?
Antla does not publish a median generation time on the listing, pricing, FAQ, or docs we opened 24 October 2026. Fast vs Pro is a dashboard switch. Unlimited Fashion lists Priority Image Gen. Headless docs suggest polling about every two seconds and picking your own UI deadline. Time it in the 7-day trial.
Does Fast vs Pro change the wait?
Yes, as a product choice: Fast is quicker, Pro is higher fidelity. No seconds are published for either. Score both on the same SKUs.
What happens to the customer’s photo?
Antla’s privacy policy states: “Images are deleted and not accessible within 72 hours.”
Does the wait include a size recommendation?
Not on Antla. Antla has no size chart or size recommendation of its own. It works alongside any Shopify size app. Looksy’s 20-second line includes a size pick. That is a different job.
Is Antla SOC 2 or ISO certified?
No. Antla is Built for Shopify certified. It is not SOC 2 or ISO certified.
Time it on five SKUs, on a phone: Antla.io/demo · Install: apps.shopify.com/antla
Related: Virtual try-on on Shopify Dawn · Try-on button placement on the PDP · Virtual try-on app for clothing · No-code virtual try-on on Shopify · Explain AI clothing previews on the PDP
About the author: Aaron leads Antla and times a try-on from upload confirm to still, not from the homepage GIF.