ChatGPT launches virtual try-on feature for viewing clothes and accessories
Read more
Olhar Digital
olhardigital.com.br

ChatGPT launches virtual try-on feature for viewing clothes and accessories

ChatGPT has received a new feature aimed at users looking for clothing and accessories online. OpenAI has started the global release of a virtual try-on feature, which allows users to demonstrate how an item of clothing might look on them using a selfie or a full-body photograph.

This new functionality is present on product pages presented by ChatGPT during shopping searches. When selecting an item or accessory, the user finds the 'Try on' button, which allows sending an image and generating a new representation with the item overlaid on the body.

Additionally, the process can be performed with items located outside of ChatGPT's direct search results. It is sufficient to send an image of the product, such as a screenshot from an online store, and ask the chatbot to show how it would look on the user.

This addition integrates a broader strategy by OpenAI, aiming to transform ChatGPT into a complete product discovery tool. In addition to virtual trying on, the user can describe the desired style and ask for suggestions of items to create a look. It is also possible to upload a photo of an outfit worn by a celebrity and ask ChatGPT to identify similar available articles for purchase.

The company has also implemented the Favorites feature, which allows users to save products found in searches to a dedicated section in the ChatGPT Library. Items can be saved individually or organized into folders for easy later access.

More information:

Despite operating as a digital try-on, OpenAI itself issues a warning that the generated images may not exactly match the real product or the user's appearance. The simulation also does not guarantee the expected size or fit of the garment. For this reason, the company advises checking the measurements provided by the seller, the product details, and the return policy before finalizing the purchase.

The technology used in these new shopping experiences is based on ChatGPT Images 2.5, a model that OpenAI claims provides more natural lighting, more detailed textures, greater ability to follow editing commands, and lower latency in image creation.

The virtual try-on marks another advancement by OpenAI in transforming ChatGPT into an online shopping portal. In March, the company had already announced a more visual product discovery experience, which included comparing options and displaying updated information directly in the conversation. Now, in addition to assisting with product location, the chatbot offers a way to preview an item on the individual before making a purchase decision.

Product results are selected autonomously by ChatGPT and, according to OpenAI, do not constitute advertising nor are they influenced by company partnerships. When there are advertisements, they are managed separately from product results.

Similar stories

Chinese startup Spirit AI develops brains for humanoid robots focusing on embodied AI
Read more
olhardigital.com.br

Chinese startup Spirit AI develops brains for humanoid robots focusing on embodied AI

The Chinese startup Spirit AI is working on developing brain systems intended for humanoid robots, allowing them to perform complex tasks based on voice commands. The company's core purpose is to advance embodied artificial intelligence, integrating machines in both production environments and commercial services.

Although progress in the physical aspects of robots is happening rapidly, software remains the main bottleneck in the sector. To overcome this barrier, Spirit AI creates advanced models capable of interpreting verbal instructions and executing actions in the real world. Gao Yang, who serves as co-founder and chief scientist of the company, highlights the current fragility of the system, stating that the brain constitutes 'the weakest link in the entire robotics stack.'

Implementation Prospects

The company plans to achieve significant advancements in industrial and commercial contexts before introducing the robots into homes. It is expected that natural language interactions will become viable by mid-2027. There is a projection that the equivalent of GPT-3.0 will be reached in the same half of 2027, enabling the user to converse with a robot in natural language, and have it execute a coherent sequence of physical actions to fulfill the request.

For training its algorithms, Spirit AI avoids relying primarily on virtual simulations, instead using approximately a thousand employees equipped with sensors to capture everyday movements. This data structure already supports practical uses, although challenges remain, such as the difficulty in perfecting fine movements, like unscrewing a bottle cap.

In terms of safety, the robots employ full-body force control and have an automatic emergency braking system when detecting any excessive force during environmental interaction. By utilizing data from real-life situations, Spirit AI seeks to transfer artificial intelligence from digital interfaces to tangible applications, beginning its journey in factories and commercial service sectors.

Popular