How to generate AI fashion models with the Photoroom API

Developers and e‑commerce teams generate AI fashion models at scale by integrating the Photoroom Image Editing API into their catalog production workflow. The AI Fashion Models (previously Virtual Model) API feature turns a clothing photo into a finished on-model shot programmatically through the same /v2/edit endpoint* as the rest of the Image Editing API, with no manual step per image. E‑commerce teams use the same request pattern to automatically produce model images across thousands of SKUs in a single workflow.

An AI fashion model is a synthetic human figure generated by artificial intelligence to wear your product in a finished image. On-model photography is any product image showing the item on a person, whether that person was photographed in a studio or generated by AI.

This matters because every fashion SKU needs on-model photography to sell, studio production cannot cover catalog volume at e‑commerce listing speed, and general-purpose AI image tools distort product details and can swap model identities between edits. Photoroom provides the AI Fashion Models API that e‑commerce teams use to generate on-model photography at catalog scale.

This guide covers how to access AI Fashion Models through the Photoroom API, what you can control in each request, how to automate production across your full catalog, and how Virtual Try-On lets shoppers see products on themselves.

Are AI fashion models available through the Photoroom API?

AI fashion models are fully available through the Photoroom Image Editing API. While the Photoroom web app interface requires a more hands-on approach to generate on-model images in batches, the API integrates with your product systems and lets engineering teams automate on-model photography at scale across your entire catalog.

The AI Fashion Models generator in the API is part of the Photoroom Plus and Enterprise plans, and it produces images continuously as new products enter your systems, with no per-batch trigger from your team:

• Developers send a flat-lay or ghost-mannequin image to the /v2/edit endpoint with the virtualModel parameter, and the API returns a finished on-model image. 

• Request parameters control the model (16 preset models), the scene (23 presets or a custom scene), the pose (12 presets), and the output size and aspect ratio. 

• The default output resolution is 1K; 2K and 4K are available on Enterprise plans.

The Photoroom API provides developers with a single endpoint for on-model generation at catalog scale. If your team needs a consistent brand model for the output, you can supply images of your brand’s model via a custom parameter instead of using presets.

The Photoroom API turns a single jewelry product photo into a finished on-model shot

How do you generate an AI fashion model programmatically?

You can generate an AI fashion model programmatically in three steps: activate the Photoroom API and get an API key, send a product image to the /v2/edit endpoint with the virtualModel parameters, and set the model, scene, and pose values in the request. The API returns a finished on-model shot from a single call, repeating the same request pattern for every product in your catalog.

1. Activate the API and get your API key

Here are the first steps for using the AI Fashion Models API feature in the Photoroom API:

• Get an API key: Sign up at the Photoroom API dashboard, activate the API, and retrieve your unique x-api-key. Create the account that owns the key using a generic company email ([email protected] rather than a personal address), so your team maintains control of the key when people change roles. You will pass the key in the x-api-key header with every request.

• Prepare your image: Use a clear, well-lit flat-lay or ghost-mannequin product image.

• Environment: Ensure you have Python installed with the requests library (pip install requests).

To test before choosing a plan, prepend sandbox_ to your key. Sandbox mode gives you up to 1,000 free calls per month with every Image Editing API feature enabled, and watermarked outputs.

2. Send your product image to the edit endpoint

Post your flat-lay or ghost-mannequin image to the /v2/edit endpoint. A flat-lay is a product photographed from above on a flat surface, and a ghost-mannequin image is a garment photographed on a mannequin and then removed, leaving the clothing with its worn, three-dimensional shape. If your product images are hosted online, you can send an imageUrl in a GET request instead of uploading the file. 

The following curl call generates an on-model shot using the avery preset model in a street scene. curl gives you a quick way to test the output from your terminal before you write any integration code.

curl --request POST \
  --url https://image-api.photoroom.com/v2/edit \
  --header 'x-api-key: YOUR_API_KEY_HERE' \
  --form imageFile=@/path/to/image.png \
  --form removeBackground=false \
  --form referenceBox=originalImage \
  --form virtualModel.mode=ai.auto \
  --form virtualModel.model.preset.name=avery \
  --form virtualModel.scene.preset.name=street \
  --form virtualModel.pose=standing \
  --output result.png

Note: The virtualModel prefix is not outdated. The API kept its original parameter names when the product was renamed to AI Fashion Models. 

When the terminal test returns your expected image, move the same request into your integration code. The following Python script sends identical parameters and saves the returned image:

import requests

response = requests.post(
    "https://image-api.photoroom.com/v2/edit",
    headers={"x-api-key": "YOUR_API_KEY_HERE"},
    data={
        "removeBackground": "false",
        "referenceBox": "originalImage",
        "virtualModel.mode": "ai.auto",
        "virtualModel.model.preset.name": "avery",
        "virtualModel.scene.preset.name": "street",
        "virtualModel.pose": "standing",
    },
    files={"imageFile": open("path/to/image.png", "rb")},
)

with open("result.png", "wb") as f:
    f.write(response.content)

Keep removeBackground=false and referenceBox=originalImage in your call. The API removes backgrounds by default, so deleting these two settings strips the generated street scene and returns the model on a transparent background. Remove them only when you want a transparent cutout. 

3. Configure the parameters

The example call produces one styling: the Avery model, in a street scene, standing. To generate a different one, change any of these values and send the call again:

• virtualModel.model.preset.name switches the model (jackson, ava).

• virtualModel.scene.preset.name switches the scene (studio, beach).

• virtualModel.pose switches the body position (seated, walkingforward).

Repeat until the output matches the look your brand needs. You’ll find the full set of models, scenes, and poses in the next section of this article.

Any parameter you leave out returns to its default, so the minimal call still works and returns the Avery model in a random scene with a random pose. 

That’s the complete integration: one key, one endpoint, and a handful of parameters your code sets per product. Photoroom turns a product photo into a finished on-model shot in one API call. What remains is deciding which model, scene, and pose fit your brand, and the Photoroom API provides several options to choose from.

A marketplace seller boots product photo rendered on a model with select scene and model pose in Photoroom

What can you control in the Photoroom API for AI fashion models?

You control five things about every AI fashion model when using the Photoroom API: the model wearing the product, the scene behind them, the pose, the output size, and the product references the generation draws from. Each is a parameter you set per request, so every product in your catalog can have its own styling or share one.

1. Preset and custom models

The model is the first decision, and you have two ways to make it:

• virtualModel.model.preset.name selects one of 16 preset models by name (Avery, Jackson, and Ava among them).

• virtualModel.model.custom uses your own brand model instead. You supply a photo of your model via imageFile in a POST request, or imageUrl if the photo is hosted, and the output renders that person wearing your product. 

A custom model keeps one identity across your entire catalog, so a buyer scrolling hundreds of your listings connects the same face to your brand across platforms.

A custom brand model in practice: upload your model's photo once, and the AI Fashion Models generator renders every product on the same person. 

2. Scenes, poses, and output size

The setting around the model works the same way as the model itself:

• virtualModel.scene.preset.name selects one of 23 preset scenes, from street and studio through beach, library, and nightlights. A custom scene upload is also available, using the same imageFile and imageUrl pattern as custom models.

• virtualModel.pose selects one of 12 poses, including standing, seated, walkingforward, and overtheshoulder.

• virtualModel.size sets the output's aspect ratio, with seven values across portrait, landscape, and square formats, defaulting to PORTRAIT_HD_3_2 (portrait at 3:2).

Resolution is independent of the aspect ratio. Outputs render at 1K by default, and 2K and 4K outputs are available on Photoroom's Enterprise plan.

Shirt product photo converted to on-model images with different models, poses, and scenes

3. Prompts and additional product images

The API also accepts two optional parameters, one for accuracy and one for style:

• virtualModel.additionalProductImages[] sends extra photos of the same product, showing different angles or details, so the generation renders the garment more accurately.

• virtualModel.prompt guides the overall style using a short text prompt, such as street style.

The following table summarizes the API request parameters and what each controls:

Read the AI Fashion Models endpoint reference for the full parameter documentation, including photos of every preset model, scene, and pose.

Photoroom makes every one of these styling decisions a parameter your code sets per product. Setting them once per catalog, instead of once per image, is what the next section covers.

How do you automate AI fashion model production across your full catalog?

You automate AI fashion model production by connecting the Photoroom API to your Digital Asset Management (DAM) system or product platform. Your DAM calls the /v2/edit endpoint for each new SKU. The endpoint applies the edits each image needs, such as background removal, Flat Lay, Ghost Mannequin, and AI Fashion Models, and returns the finished on-model shot in one API flow.

Automation at catalog scale also removes the human image reviewer from the processing loop. Three parts of the Photoroom workflow take over that job:

• Consistent model identity: Preset models return the same identity on every call, so a catalog styled on a specific model preset reads as one photoshoot no matter when the API generated each image. You can achieve the same consistency with your custom brand model. Upload their photo once, and every on-model shot across thousands of SKUs renders the same person.

• Automated fidelity checks: Photoroom’s Visual QA scores each output against the original image for fidelity using vertical rater models; Fashion Rater for fashion catalogs and Food Rater for food platforms. When a score falls below the threshold, Visual QA locates the mismatch and fixes it automatically.

• Automated accuracy correction: Photoroom's Fidelity Layer compares the generated image to the original product, identifies any fidelity issues, localizes the problematic areas, and uses that analysis to preserve product accuracy, raising accuracy above every base frontier model in the Photoroom Fidelity Benchmark.

For enterprise teams, Photoroom’s Enterprise Guarantee adds a quality commitment to your workflow. The Enterprise Guarantee for AI visuals lets you set your product-fidelity criteria upfront, and the platform re-generates any image that fails your contractual criteria for you, so you pay only for the outputs you accept. Photoroom tests your sample images through the platform before any contract is signed, and the pass/fail criteria are agreed only once those results pass on your actual catalog. 

Global sporting goods retailer Decathlon uses Photoroom's API to standardize product images at scale, with 500 product categories and 150 packshot guidelines to enforce. The result? 99% of products pass automated quality checks against Decathlon's brand standards without manual review.

Photoroom keeps one model identity across your catalog, checks every output before it ships, and puts pass/fail terms in the contract. The same technology now extends to shoppers who want to see products on themselves before buying.

How to integrate Virtual Try-On for shoppers to visualize products on themselves?

The Virtual Try-On feature in the Photoroom API lets shoppers see products on themselves in your own app or website, using the same AI Fashion Models technology that fills your catalog.

Here’s how Photoroom’s Virtual Try-On API feature works:

• Integrate once. Your engineering team integrates the Photoroom API by following the Virtual Try-On documentation.

• The shopper uploads a photo of themself in your app or your website.

• The photo becomes a custom AI fashion model. The API turns that photo into the shopper's own model.

• The product renders on the shopper. The API returns the selected item shown on their body, and the shopper never leaves your store.

Shoppers who can visualize the product on their own bodies buy with more confidence and return fewer items due to size or taste objections, making Virtual Try-On a good fit for enterprise retailers, marketplaces, and fashion brands with the developer resources to integrate the API and a catalog where fit drives purchase decisions.

If you’re considering building a try-on technology in-house, consider that developing the technology internally creates more than a compute and engineering-cost challenge: it also makes your marketplace or retail platform responsible for protecting sensitive shopper images, managing consent and retention, and maintaining user trust when results are inaccurate. Issues such as incorrect garment shape or body-type representation can damage marketplace credibility. 

Photoroom addresses these challenges by providing the Virtual Try-On solution through a production-ready API that maintains product accuracy and meets enterprise-grade security standards. The platform encrypts data in transit with TLS 1.2 or higher and at rest at the infrastructure layer, and deletes operational logs after 15 days. These provisions reduce the cost, operational complexity, and trust risks associated with developing try-on technology from scratch.

Shopper photo converted into an on-body product render with Photoroom Virtual Try-On

Photoroom API pricing

Photoroom offers flexible API pricing based on your needs. AI Fashion Models is included in the Plus and Enterprise plans:

View full pricing details →

Photoroom is the AI product photography platform for fashion e‑commerce at scale, providing an API that turns a product photo into a finished on-model shot in one call, catalog automation that keeps one model identity across thousands of SKUs, a fidelity system with pass/fail terms written into the contract, and Virtual Try-On that lets shoppers see products on themselves before they buy.

*The /v2/edit endpoint is the web address your system sends images to when calling the Photoroom Image Editing API. Your code sends an image and its parameters to https://image-api.photoroom.com/v2/edit, and the API returns the edited result; every edit in this guide goes through this single endpoint.