OpenAI Adds Virtual Try-On and Document Scanning to ChatGPT
ChatGPT can now generate clothing try-on previews from a saved reference photo and turn a sequence of mobile camera captures into one PDF.
Edited by Tyronne Panaino
OpenAI added two consumer-facing input and image features to ChatGPT on October 1. Product listings for clothing and accessories can now offer a virtual try-on generated from a selfie, while ChatGPT's mobile camera can capture several document pages and combine them into one PDF for a conversation.
The shopping feature is available on mobile and web, according to the official ChatGPT release notes. Document scanning is rolling out on iOS. Together, the changes affect people using ChatGPT to compare products visually and people who need to bring paper notes or documents into a chat without assembling a PDF in another app first.
Virtual try-on adds a saved reference photo
A Try on button can appear on ChatGPT product listings for clothes and accessories. The user takes or uploads a selfie, and ChatGPT Images uses that reference to generate a visualization of how an item could look. OpenAI says the reference photo is saved for future try-ons. Users can change or delete it under Personalization settings.
That persistence is an important part of the feature, not a background detail. Reusing a reference photo can make repeated shopping comparisons faster, while saving a personal image also gives users a reason to review whether the stored photo is still wanted. The release note identifies the deletion control, but it does not provide a technical account of retention periods, image processing, regional storage or model accuracy.
ChatGPT also lets users save products to Favorites or create a folder for finds in the Library. That turns the update into more than a single image-generation action: products can be revisited and organized inside the same service. OpenAI does not say in the fetched release note which merchants or product categories support the button, how broadly it is already visible, or whether generated previews are available in every market.
The camera can assemble multi-page PDFs
The second change targets document capture. On a phone, a user can take several page images in sequence and ChatGPT combines them into a single PDF that is ready to upload to the conversation. OpenAI says the flow is accessed from the camera's three-dot menu and is rolling out on iOS.
The practical delta is workflow compression. A handwritten set of notes, a printed form or a short packet can enter ChatGPT as one document without a separate scanning application creating the PDF first. The release note establishes the capture and combination steps; it does not claim optical-character-recognition accuracy, automatic page cleanup, maximum page counts or Android availability. Those details should not be inferred from the launch description.
What users should verify
A generated clothing preview is not the same thing as a verified statement about size, fit, fabric or color. OpenAI describes a way to see how an item could look, but the fetched evidence supplies no measurement method, merchant validation or independent accuracy test. People making a purchase should therefore treat the output as a visualization rather than proof that a garment will fit.
Document users should similarly check the resulting PDF before relying on it. The release note does not describe how the scanner handles faint text, glare, rotated pages, handwriting or missing edges. A convenient upload path can reduce manual steps without guaranteeing that every captured page is readable or complete.
What to watch next
The next useful checkpoints are wider availability details, explicit handling rules for saved reference photos, support for document scanning beyond iOS and evidence about the quality limits of both features. Independent testing would also be needed to compare generated previews with real garments or to measure the scanner across difficult documents.
Status
Confirmed. OpenAI's official October 1 changelog establishes both features and their stated platform availability. Internal confidence is medium because the evidence is one first-party source and AI News Bank did not independently test rollout coverage, image accuracy, scan quality or data handling.
Sources
Update note: Last reviewed 2026-10-02. We will revise this post if OpenAI changes availability, reference-photo controls or document-scanning support.
Sources
- OpenAI — ChatGPT release notes — official
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.