Google Virtual Try-On: Three Outfit Examples

A try-on starts with two images whose roles are clear: a person photo and a garment photo. Google Virtual Try-On has no prompt input. Choose images that make the person's pose, garment outline and occlusions easy to see.

TL;DR

Each result costs $0.1075; see pricing and budgets. Send person_image_url, product_image_url and num_images: 1. The three examples below cover a striped top, a satin dress and an outdoor shirt. Each set is ordered as person, garment, then result.

Example one: a striped T-shirt with jeans

Upload images to try it →; no prompt required.

Upload images to try it →; no prompt required.

Upload images to try it →; no prompt required.

The top becomes a red-and-white striped T-shirt. The gray background, jeans and white shoes remain recognizable. Stripes curve across the torso and sleeves, and the hem position changes. For a product image, inspect the neckline, stripe direction and transition at the waist as well as the color.

This pairing is a useful starting point: the person faces forward with the upper body visible, and the garment photo includes the sleeves and hem. To use the same pairing, place person A's source in person_image_url, the striped flat lay in product_image_url, and set num_images: 1. Feeding the already dressed result back into the person field changes the input rather than providing another candidate under the same conditions.

Judge the stripes and neckline against the product first, then inspect garment edges near the arms and the waistband transition. For an outfit illustration, overall color and styling may suffice. If stripe placement or hem length is a selling point, compare those areas with the garment photo. Body proportions and folds cannot establish that a size fits or replace information about the fabric's actual stretch.

Example two: a satin dress and changed accessories

Upload images to try it →; no prompt required.

Upload images to try it →; no prompt required.

Upload images to try it →; no prompt required.

The green color, straps and draped neckline are visible in the result, replacing the original top and jeans with a long dress. The white shoes also become heeled sandals, and folds near the right hand resemble a pocket opening. This set can help assess an overall outfit, but a brief that allows only garment changes requires a separate check of the shoes and the area around the hand.

Keep person B's original black-top-and-jeans photo and pair it with the independent green dress image. A long dress replaces more of the outfit than a short top, so acceptance should include the hem, ankles and shoes rather than only a crop around the straps. The original outfit helps identify changes, but the public input has no option to lock a particular accessory.

Decide whether the intended use allows different shoes. For exploring evening-outfit directions, the dress and sandals can be a candidate. If the brief requires the original white shoes, this result does not satisfy it. Folds near the hand are not evidence that the product has a pocket. Confirm its structure from the garment image or product information before writing selling points based on a generated shape.

Example three: an Oxford shirt in a street scene

Upload images to try it →; no prompt required.

Upload images to try it →; no prompt required.

Upload images to try it →; no prompt required.

The street, beige trousers and standing pose largely carry through while the white T-shirt becomes a light-blue shirt. Sleeves that extend fully in the flat lay are rolled toward the forearms in the result, and the hem is tucked into the trousers. If the brief needs a particular sleeve length or an unrolled look, check the result against that requirement rather than assuming every detail carries over unchanged.

Keep the person and product roles separate: the street photo supplies the pose and setting, while the independent shirt image supplies the garment appearance. Use person C and the blue-shirt addresses for this pairing. To see the shirt on someone else, establish another combination. Changing only the input you are comparing makes the selection decision easier to explain.

Rolled sleeves and a tucked hem create a particular styling choice while hiding the product's original cuff and hem lengths. That may fit a street-outfit illustration. A product page meant to show long sleeves or a loose hem needs a result or product photo that meets those requirements. There is no prompt for unrolled sleeves or text parameter to force a wearing style. More candidates provide choices without guaranteeing a specific detail.

Select inputs and review the result

Use a single person, preferably facing forward with the body clearly visible. Choose a clean garment photo that includes the neckline, sleeves and hem. Generate one candidate, then inspect the garment, hand occlusions, accessories and background. Compare specific product details with the original when they matter to the brief.

None of these examples needs a prompt. Open the entry below an image and upload a person photo and garment photo; the integration page explains the API fields. These images show outfit appearance and do not replace sizing measurements.

Apply the examples to your own selection

Choose the observation task closest to your needs: stripes and waistband transitions in the top, broad outfit replacement and accessories in the dress, or styling and setting in the street shirt. Record that pairing, then replace inputs with your own images one at a time. Each change has a purpose; changing the person, garment and setting together makes differences harder to explain.

These examples show existing inputs and results. They cannot establish a success rate for a garment category or the number of candidates it needs. Begin with one request for each necessary combination and review it against product requirements. If you want more choices, retain the inputs and request another candidate. If the garment photo is missing an edge, return to preparation. Include new requests in the budget; additional candidates may still miss a required detail.

Record outfit usefulness separately from product-detail usefulness. An outfit illustration may allow rolled sleeves, a tucked hem or changed shoes; product imagery must describe the item accurately. Folds resembling a pocket or clothing looking close-fitting are not product specifications. The examples identify areas to inspect, while acceptance still requires your product information.

Scope of request reproduction

The request contains the model name, two image fields and num_images: 1. The integration page has complete cURL and Python examples. These cases have no prompt to copy; the entry below each image accepts person and garment uploads. Resubmitting the same images reuses input conditions without promising a pixel-identical result.

Keeping source files helps explain whether a difference comes from the garment, the person's pose or the generated image. A result without its inputs makes product fidelity harder to assess. Archive the selected version separately so it is not later mistaken for an original product photo.

FAQ

Why are there no prompt recipes?

The model has no prompt parameter. The person and garment images describe the task, so improving the input starts with image selection.

How do the two images differ?

The person photo shows who will wear the item. The garment photo shows the product. Their fields are person_image_url and product_image_url respectively.

Does only the clothing change?

The dress example also changes the shoes, and the shirt example changes how the sleeves are worn. Review areas outside the garment too.

How can I get several candidates for the same outfit?

Use num_images: 1 and submit separate requests with the same inputs, then compare the returned candidates.

Further reading

Pricing and budgets · API integration · Examples and input selection · Google Virtual Try-On