“Crazy”: Openai offers original GPT-4O images and is already users

We are going out on the first anniversary since Openai released the first “OMNI” or MS, GPT-4O in May 2024, but this old willingness still has some tricks.

Case at a point, today Openai has finally operated the capabilities of original multimedia images From GPT-4O for successful Chatbot Chatgpt users at Plus, Pro, TEAM, and Ertage Usage, although the company said it will soon provide institutions, EDU, and through its API’s application program (API).

Unlike the previous artificial intelligence image model available in ChatGPT – Openai’s Dall-E3The classic proliferation transformer model that was trained to rebuild images from text claims by removing noise from pixels – this new photo generator This is part of the same model that spit the text and the symbol, where Openai trained the entire form to understand all these forms of media simultaneously.

President Openai Greg Brockman A long time ago, she inspected this original GPT-4O again in May 2024, but for reasons that are still unknown publicly, the company has so far clung to it-in the wake of the general version of what many Power users saw as a similar feature of Google Ai Studio with the Gemini 2 Flash experimental model.

This has resulted in a high -quality photo generator that results in more vibrant images and an accurate, baked text, and he is already impressive of users – one of them calls for quality.crazy“

In the same manner (intended from the pun), Openai did not specifically say about the capabilities of generating GPT-4O images that were trained-and in view of the history of the company and service providers for models, it is possible that it will include many artistic works that have been embodied from the web, which are assumed to be roasted, which is likely to be angry with artists behind them.

Bring the generation of photos to ChatGPT and Sora

Openai has long aimed to make images the essential capacity of its AI models. Using GPT-4O, users can now create direct images in ChatGPT, improve them through conversation and adjust details while flying.

The model also merges in Sora, a platform or Openai to generate video, which expands multimedia capabilities.

In an advertisement, Openai confirmed that the generation of GPT-4o photos is designed for:

• Submit the text accurately inside the images, allowing the creation of signs, menus, invitations and graphs.
• Follow the complex claims accurately, maintain high accuracy even in detailed compositions.
• Based on the previous images and text, and guarantee visual consistency through multiple reactions.
• Supporting various artistic patterns, from light realism to touched illustrations.

Users can describe a picture in Chatgpt, and determine the details such as the percentage of width to height, color systems (hexagonal codes), or transparency, and will be created within one minute.

The independent artificial intelligence advisor Alli K. Miller wrote on X, it is.A huge leap in the generation of the textAnd “Best” Form of Photo Generation of the Artificial Intelligence you have seen.

Main capabilities and cases of use

GPT-4O is designed to make images generate not amazing visual but also practical. Some main applications include:

• Design and brands – Create logos, stickers and ads with a delicate text.
• Education and perception – create scientific plans, graphs, and historical pictures of learning.
• The development of the game – maintaining the consistency of the character through different design repetitions.
• Marketing and Content Creation – Production of Social Information Assets, Events Calls, and Digital Claiming Features designed to meet the brand needs.

How to improve GPT-4O obstetric images on Dall-E

According to the official Openai thread on X, GPT-4O offers many improvements to previous models:

• Better text integration: Unlike the past artificial intelligence models that struggle with a readable text in a good position, GPT-4O can now include words accurately inside the pictures.
• Understanding augmented context: GPT-4O works to take advantage of the chat record, allowing users to improve images interactively and maintain cohesion across multiple generations.
• Improving multi -object: Although previous models have had difficulty putting many distinct objects in a scene, GPT-4O can now handle up to 10-20 organisms at one time.
• Multi -use adaptation: The model can generate or convert images into a variety of patterns, from hand -drawn drawings to high -resolution realism.

Restrictions

Despite its progress, GPT-4O still has some known challenges:

• Economy issues: Large pictures, like stickers, may sometimes be tightly bolied.
• The accuracy of the text in non -Latin text programs: Some non -English letters may not present properly.
• Keep details in the small text: A very detailed or small text may lose clarity.
• The accuracy of the liberation: Modification of specific parts of the image may not affect other elements.

Openai is actively addressing these issues through continuous model improvements.

Safety measures and signs

As part of Openai’s commitment to the official AI development, all images created by GPT-4o C2PA data, allowing users to verify the origin of artificial intelligence.

Moreover, Openai has built an internal search tool to help discover images created by artificial intelligence.

There are strict guarantees to prevent harmful content and prevent misuse, such as prohibiting explicit, deceptive or harmful images.

Openai also guarantees that images that feature real people are subject to increasing restrictions.

Openai Sam Altman The version is a “high water sign for creative freedom”, with a focus that users will be able to create a wide range of visual images, with Openai monitoring and improving its approach based on the use of the real world.

Since the images created by artificial intelligence become more accurate and accessible, GPT-4O represents an important step forward in making the text generating a prevailing tool for communication, creativity and productivity.