'Insane': Openai introduces GPT-4o Native Image Generation and it's wowing users

We've been coming to a one-year anniversary since Openai released its first “Omni” or multimodal model, the GPT-4O back in May 2024, but the old standby still has some tricks on its sleeve.

Case-in-point, now Openai finally opened the native capabilities of the multimodal image generation GPT-4O for Hit Chatbot ChatGPT users in Plus, Pro, Team, and Free Tiers, although the company said it is also available for enterprise, EDU, and through the Application Programming (API) interface.

Unlike the previous generative AI image model available in ChatGPT – Openai's Dall-E 3.

Openai President Greg Brockman GPT-4O's native capabilities have long been previewed in May 2024, but for reasons remaining unknown to the public, the company has been held to this day-followed by the public's release of what many AI users have seen as a similar feature from Google Ai Studio with the Gemini Experimental Model 2 Flash.

This has resulted in a higher quality image generator that produces more living images and accurately baked text, and it impresses users – one of them calls quality “crazy. “

With the same token (Pun intended), Openai still didn't say what the GPT-4O generation generations were trained-and given the company's history and other model providers, likely to include many artwork scraping from the web, some of which may have copyright, which is likely to upset the artists behind them.

Bringing a generation of image to Chatgpt and Sora

Openai has long been aiming to create an image generation of a basic ability of its AI models. With GPT-4O, users can now generate images directly on ChatGPT, refining them by talking and fixing details quickly.

The model also includes the Sora, OpenAi generation platform, further expansion of multimodal capabilities.

In an announcement in X, Openai confirmed that the GPT-4O image generation was designed to:

• Accurately render text within images, allowing the creation of signs, menus, invitation, and infographics.
• Follow complex signals with accuracy, maintaining high honesty even in detailed compositions.
• Develop in previous images and texts, ensuring visual consistency in many relationships.
• Support different artistic styles, from photorealism to stylish drawings.

Users can describe an image in ChatGPT, specifying details such as aspects of aspects, color schemes (hex codes), or transparency, and the GPT-4O will develop it within a minute.

As independent AI consultant Allie K. Miller wrote to X, it was a “Huge jumping to the generation of text“And” the best “model of the AI ​​image generation he has seen.

Basic capabilities and use cases

The GPT-4O is designed to make the image generation not only visually stunning but also practical. Some of the major applications include:

• Design and Brand – Develop logos, posters, and commercials with accurate texture.
• Education and Visualization -Create diagrams of scientific, infographics, and historical images for learning.
• Game Development -Maintain character consistency in different design iterations.
• Creating marketing and content – make social media assets, event invitations, and digital drawings in accordance with brand needs.

How GPT-4O improves generative images in dall-e

According to the official Openai thread in X, GPT-4O has introduced some improvements to previous models:

• Better Text Integration: Unlike previous AI models that fought the readable, properly placed text, the GPT-4O can now accurately embed words within the images.
• Improved contextual understanding: The GPT-4O uses chat history, which allows users to refine images interactive and maintain unity for many generations.
• Improved Multi-Object binding: While previous models have been struggling to position many unique objects in one scene, the GPT-4O can now be handled up to 10-20 objects at once.
• Versatile style adaptation: The model can produce or change images in different styles, from drawn sketches to high-resolution photorealism.

Limitations

Despite its advancements, the GPT-4O still has known challenges:

• Crop Issues: Large images, such as posters, can sometimes be tightly tight.
• Accuracy of text in non -Latin scripts: Some non-English characters may not render correctly.
• Detail of maintenance in small text: High detailed or small text may lose clarity.
• Editing accuracy: Changing specific parts of an image may not accidentally affect the other elements.

Openai is actively addressing these issues through continuous model refinement.

Safety and Labeling Proposals

As part of the promise of Openai to the AI's responsible development, all images generated by the GPT-4O include the C2PA metadata, allowing users to prove their AI source.

Moreover, Openai has built an internal search tool to help detect images generated by AI.

Strict care is in place to block harmful content and prevent misuse, such as ban on explicit, deceptive, or harmful imagination.

Openai also ensures that the pictures featuring real people are subject to increased restrictions.

Openai CEO described Sam Altman The release as a “new high water mark for creative freedom”, emphasizing that users will create a wide range of visuals, with Openai that observes and refined the approach based on real-world use.

While the images generated by the AI-generated become more accurate and accessible, the GPT-4O represents a significant step forward in making a text-to-image generation a major tool for communication, creativity, and productivity.