ChatGPT Enhances its Image-Generation Capabilities

During a livestream on Tuesday, OpenAI CEO Sam Altman announced the first major update to ChatGPT’s image-generation capabilities in over a year.

Now, ChatGPT can use the GPT-4o model to create and edit images directly. While GPT-4o has long powered OpenAI’s chatbot, it previously handled only text-based tasks.

Altman stated that GPT-4o’s image-generation feature is now available in ChatGPT and Sora, OpenAI’s AI video tool, for subscribers of the company’s $200-a-month Pro plan. The feature will soon expand to Plus and free-tier ChatGPT users, as well as developers using OpenAI’s API.

Enhanced Image Precision and Editing Capabilities

http://DALL-E 3GPT-4o takes slightly longer to generate images than DALL-E 3, the model it replaces, but OpenAI claims it produces more detailed and accurate visuals. It can also edit existing images, including those featuring people, by altering elements or “inpainting” details like foreground and background objects.

To develop this feature, OpenAI told the Wall Street Journal that it trained GPT-4o using publicly available data and proprietary content from partnerships with companies like Shutterstock.

Generative AI companies often treat their training data as a key competitive asset, keeping details tightly guarded. Additionally, concerns over intellectual property disputes serve as another reason for companies to limit disclosures.

“We respect artists’ rights in how we generate outputs and have policies in place to prevent the creation of images that closely replicate the work of living artists,” OpenAI COO Brad Lightcap told the Wall Street Journal.

Creator Control and Data Privacy Measures

OpenAI provides an opt-out form that lets creators request the removal of their works from its training datasets. The company also states that it honors requests to block its web-scraping bots from gathering training data, including images, from websites.

The upgraded image-generation feature in ChatGPT comes shortly after Google introduced experimental native image output in its Gemini 2.0 Flash model. The feature quickly gained attention on social media, though not all of it was positive. Gemini 2.0 Flash’s image tool had minimal safeguards, enabling users to remove watermarks and generate images featuring copyrighted characters.

This article was updated at 12 p.m. PT to include OpenAI’s statement to the Wall Street Journal regarding GPT-4o’s training data.

Read the original article on: TechCrunch

Read more: ChatGPT Doubled Weekly Users in Under Six Months Due to Updates