OpenAI’s shopping help describes virtual try-on using ChatGPT Images. The result is a generated visual preview, so it should be distinguished from a measurement of fit.
ChatGPT virtual try-on workflow
The help page describes selecting Try on from a clothing or accessory listing, then taking or uploading a selfie. It also describes asking about an uploaded garment image in a conversation. This is the consumer workflow; it should not be confused with configuring an image-generation API.
https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search
Reference photos and fit limitations
OpenAI says reference photos are saved for reuse and can be changed or deleted in Settings, under Personalization and Reference photos. It warns that previews can differ from the product or the person and do not guarantee size or fit. Check the merchant’s measurements and returns information when making a purchase.
https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search
Developer guidance has a different role
The separate OpenAI cookbook discusses image-generation prompts for virtual try-on. It recommends explicit constraints to preserve identity and scene details while changing garments, with attention to draping, occlusion and lighting. Those are instructions for a generation workflow, not a promise that the consumer product always preserves every detail or predicts garment measurements.
https://developers.openai.com/cookbook/examples/multimodal/image-gen-models-prompting-guide
Sources and further reading
https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search
https://developers.openai.com/cookbook/examples/multimodal/image-gen-models-prompting-guide