Shopify
WooCommerceWe sent the same 10 shopper photos and products to 10 virtual try-on APIs. Default settings, one run each, no retouching. Here is what came back.
Disclosure: Genlook makes one of the ten APIs. Every result is shown below so you can judge for yourself.
Benchmark Results
Green marks the winner in each category.
| API | Price per try-on | Median time | Garment fidelity | Products covered | Unwanted changes |
|---|---|---|---|---|---|
| Genlook | $0.02 | 9.8s | Excellent | All, one endpoint | none |
| Google Vertex AI | $0.048 at volume ($0.06) | 11.9s | Excellent (clothing) | Clothing only | none |
| Gemini 3.1 Flash Image | $0.068 | 11.6s | Very good | Any, via your prompt | none |
| GPT Image 2 | $0.038 average (medium quality) | 20.5s | Very good | Any, via your prompt | 1: changed shoes |
| FASHN Try-On Max | $0.15 (1K, balanced) | 25.0s | Very good | Clothing, shoes, hats, jewelry | none |
| Seedream 5.0 Pro | $0.093 | 35.3s | Good, paler colours | Any, via your prompt | none |
| Pixelcut | $0.10 (+$0.05 extraction) | 51.9s | Good | Clothing, best on tops | 1: coat lengthened |
| FLUX.2 Pro | $0.045+ | 19.1s | Mixed | Any, via your prompt | 2: mirrored text, invented sleeves |
| FASHN v1.6 | $0.049 at volume ($0.075) | 13.1s | Mixed | Clothing only | 2: replaced trousers, drape artifact |
| FLUX VTO (Black Forest Labs) | ~$0.04 on hosts | 19.4s | Mixed | Clothing only | 2: kept old trousers, garment not applied |
Garment fidelity is how faithfully the product is rendered (colour, print, texture, cut). Unwanted changes counts results where something unrelated to the product changed, like shoes or invented details. Genlook's time is its production median; the others were timed by our script. General image models got one shared try-on prompt.
Top 5 Virtual Try-On APIs
1. Genlook
The most faithful colours and textures in our test, the fastest median time (9.8s) and the lowest price ($0.02). It is also the only unified try-on endpoint in the group: the same call handles clothing, shoes, glasses, jewelry and hats, with no prompt or category to manage. Vertex matches it on clothing but covers nothing else, at three times the price. Docs, Get started.
2. Google Vertex AI Virtual Try-On
Excellent on clothing: a dedicated model with the cleanest result set, it never changed anything but the garment. The limits are scope and cost: it only handles garments (no shoes, glasses or jewelry), costs $0.06 per image, and needs a Google Cloud project. Docs.
3. Gemini 3.1 Flash Image
A general image editor, not a try-on product, yet with a one-line prompt it matched Vertex on quality and speed (11.6s) at $0.068. The trade-off: you write and maintain the prompt, and migrate when Google retires a version. Docs.
4. GPT Image 2
Strong garment rendering, and the cheapest of the general models at medium quality ($0.038 on average). Twice as slow as Gemini (20.5s), and it restyled the shopper's shoes once. Docs.
5. FASHN Try-On Max
The best lace and fine detail after Genlook, but the most expensive ($0.15 per image at 1K) and slow (25s). It sometimes picks the wrong item from on-model photos unless you set the garment category. Docs, Genlook vs FASHN.
The other five (Seedream 5.0 Pro, Pixelcut, FLUX.2 Pro, FASHN v1.6 and FLUX VTO) all produced usable images, but were slower, paler or made more unwanted changes.
Key Findings
- Colour and texture separate the leaders. Everyone gets the garment roughly right. Few get the exact tone, print and lace.
- Failures are collateral damage. Swapped shoes, replaced trousers, invented sleeves. Test what should not change.
- On-model product photos are ambiguous. Shown a model in jeans and a t-shirt, most APIs dress the shopper in both. Use packshots when you can.
Results by Garment
Trench coat (packshot)
Genlook matches the ecru tone, gold buttons and knee length most closely. FASHN v1.6 replaced the jeans with white trousers; Pixelcut stretched the coat to the ankle.
Football shirt with sponsor text
Text held up almost everywhere. FLUX.2 Pro mirrored the sponsor name; GPT Image 2 changed the shoes.
Silk saree
A hard drape, handled well by nine of ten. FASHN v1.6 left a panel of fabric floating beside the body.
Printed silk midi dress
The print survived everywhere. FLUX VTO layered the dress over the shopper's trousers.
Lace wedding gown
Genlook and FASHN Try-On Max render the lace bodice most faithfully. FLUX.2 Pro invented long sleeves.
Abaya
Genlook, GPT Image 2 and FLUX.2 Pro dressed the full look from the product photo, cream hijab included. FLUX VTO only changed the top half.
Jeans, shot with a t-shirt
Most APIs dressed the full outfit, with the best denim wash on Genlook. FASHN Try-On Max and FLUX VTO picked the t-shirt or changed nothing.
Other Try-On APIs
Not in our test, prices from each vendor's site in September 2026.
| Vendor | Price per try-on | Notes |
|---|---|---|
| Pruna P-Image-Try-On | $0.015 first garment | Up to 11 garments per call |
| Alibaba OutfitAnyone | ~$0.03 to $0.07 | Beijing region only |
| Magic Hour | ~$0.025 (annual plans) | Resolution capped by plan |
| FitRoom | ~$0.09 | Self-serve |
| Photoroom | $0.10 | 1K output self-serve |
| Perfect Corp YouCam | ~$0.10 to $0.11 | Also shoes, bags, jewelry |
| Kling AI | $0.112 | $350 standard pack |
| Claid | $0.118 | Up to 5 garments |
| Aiuta | Not public | SKU catalog based; powers ASOS |
| WeShop | Enterprise, $457/month | Try-on gated to top plan |
Open source: most try-on models (IDM-VTON, CatVTON, OOTDiffusion) are licensed for research only. FASHN VTON v1.5 and Leffa allow commercial use.
How to Choose
- Test your five hardest products, not a demo gallery.
- Check what should not change: face, body, shoes.
- Count the whole bill: retries, storage and the wait your shoppers sit through.
- Prompted models like Gemini work well, but you maintain the prompt.
Genlook costs $0.02 per try-on, with three free credits to start. Read the docs and test it on your own catalog.
FAQ
Questions, answered.
What is the best virtual try-on API in 2026?↓
In our 10-pair test, Genlook had the best garment fidelity, the fastest median time (9.8 seconds), the lowest price ($0.02 per try-on) and the only unified endpoint covering every product type. Google Vertex AI was excellent on clothing only; Gemini 3.1 Flash Image, GPT Image 2 and FASHN Try-On Max were also very good.
How much does a virtual try-on API cost?↓
From $0.015 to over $0.30 per image. Typical prices: Genlook $0.02, Google Vertex AI $0.06, Gemini about $0.07, FASHN $0.075, Pixelcut and Photoroom $0.10.
Can Gemini or GPT Image do virtual try-on?↓
Yes, with a prompt that passes both photos. Both performed very well in our test; GPT Image 2 changed the shopper's shoes once.