OpenAI Launches Virtual Try-On in ChatGPT With New Images 2.5 Model
ChatGPT now generates virtual try-on photos from user-uploaded selfies using the Images 2.5 model, copying Google's 2025 feature as OpenAI builds out shopping revenue.
Updated
Why it matters
- ChatGPT's new Virtual Try On feature generates photos of users wearing clothes from an uploaded selfie and full-body image, using the ChatGPT Images 2.5 model released in September.
- The feature mirrors a virtual try-on capability Google added to Search in 2025, and works with any uploaded clothing screenshot, not just products surfaced in chat.
- Uploaded images are used for training by default unless users opt out; OpenAI also added a Library for saving products, alongside existing Instant Checkout fees and in-chat ads.
ChatGPT can now generate virtual try-on photos, letting users see themselves in clothes by uploading a selfie and a full-body picture. OpenAI built the feature on the ChatGPT Images 2.5 model it released in September, and it arrives alongside a new option to save products to a Library inside the app.
The mechanics are straightforward. Anywhere a product listing for clothing appears in a chat, users tap on the garment, tap "Try on," and upload their photos. ChatGPT then generates an image of them wearing the item. The feature is not limited to products the chatbot surfaces: according to the source, any screenshot of clothing a user uploads could theoretically be used to generate a try-on image.
A familiar playbook
The concept is identical to a feature Google added to Search in 2025, which similarly lets users upload photos and see themselves in a variety of clothes. OpenAI is following an established pattern rather than inventing a new category. That framing matters for expectations: virtual try-on has already been tested at scale by a company with far larger shopping traffic, and OpenAI is now betting the same capability can drive engagement inside a conversational interface.
The privacy catch
There is a familiar caution attached. By default, any images uploaded from a personal ChatGPT account can be used for training unless the user opts out. A feature that invites people to upload selfies and full-body photos widens the surface of that data collection. Users who want to try on clothes without contributing to model training need to change their settings before uploading.
Why OpenAI keeps pushing shopping
OpenAI is not an online marketplace on the scale of Amazon. But the company has started taking a cut of certain transactions completed in ChatGPT through its Instant Checkout feature. It also offers ads inside chats that merchants can use to advertise products and services.
Neither revenue source seems likely to account for the exorbitant amount of money OpenAI spends training and running its AI. Even so, the consistent rollout of new shopping features over the last year — Instant Checkout, ads, and now virtual try-on plus a saved-products Library — signals that OpenAI believes commerce inside ChatGPT could make a financial difference some day.
The stakes are real. OpenAI's compute costs are among the largest in the industry, and advertising- and transaction-based revenue is the proven model that funds companies like Google. Virtual try-on is a small feature, but it sits squarely in a strategy to turn ChatGPT from a subscription and API business into something closer to a commerce platform.
For now, the feature is live wherever clothing listings appear in ChatGPT conversations, with generated images produced by the ChatGPT Images 2.5 model and saved items stored in the app's Library for later revisiting.
Original: openai.com
More from Rebecca Stone
Show full bio
Correspondent covering consumer brands and retail at AI In Context.
177 articles
Related articles
- OpenAI Launches Virtual Try-On and Favorites in ChatGPT
- OpenAI Ships gpt-image-1 to API After 700M-Image Debut
- OpenAI Ships ChatGPT Images 2.5 With Sketch Tool and 50% Faster Generation
- OpenAI Rebuilds ChatGPT Shopping Around Visual Discovery and ACP
- OpenAI opens ChatGPT app submissions and launches in-app directory