GPT-4o: Stunning images on ChatGPT… but at what cost? 🚀🌍

Cliquez ici pour lire en français

Since its launch, ChatGPT has established itself as an essential tool for generating text fluently and relevantly. Today, OpenAI takes another step forward by integrating image generation directly into ChatGPT with its advanced model, GPT-4o. This update promises to transform how we interact with AI in visual creation… but it also raises environmental concerns.

GPT-4o: the new visual engine of ChatGPT 🎨

Until now, image creation via ChatGPT relied on DALL-E, a powerful yet sometimes limited tool. With the arrival of GPT-4o, an « omnimodal » model capable of processing text, images, sound, and video, image generation becomes more precise and detailed. Users can now obtain high-quality visuals directly within the ChatGPT interface without relying on external tools.

Image of Karl Marx generated with ChatGPT

More precise and detailed images 🖼️

One of GPT-4o’s major advancements lies in its ability to improve the rendering of text embedded in images. Unlike previous models, GPT-4o adopts an autoregressive approach, generating images from left to right and top to bottom, allowing for better detail management and increased text readability. Moreover, it can handle up to 20 distinct objects in a single image, offering unprecedented visual richness.

What types of visuals can be created? 🎨

Thanks to GPT-4o, it is now possible to generate a wide range of illustrations, from photorealism to more artistic styles like sketches, watercolor, or animation inspired by Ghibli-style aesthetics. This model also manages complex compositions with multiple distinct objects and better representation of faces and hands.

Image generated with ChatGPT

With GPT-4o, users can create portraits, detailed landscapes, conceptual images for advertising, or even science fiction scenes. It can even integrate text into images with improved readability, opening up possibilities for more precise posters and memes.

Image generated with ChatGPT

These advancements open the door to numerous practical applications across various fields:

• Marketing & Communication: Rapid image generation for advertising campaigns or social media posts.
• Education: Creation of educational visual materials, such as illustrated diagrams or infographics.
• Video Games & Design: Prototyping environments and characters before detailed design.
• Media & Journalism: Illustrating articles without relying on traditional image banks.

Image generated with ChatGPT

Image generated with ChatGPT

Image generated with ChatGPT

These new capabilities make image creation more flexible and faster, though their use must be carefully considered due to the associated environmental impact.

A gradual rollout… and already overloaded 🌍

Currently, this feature is available for ChatGPT Plus, Pro, and Team subscribers. However, due to higher-than-expected demand, the rollout for free users has been temporarily delayed. Sam Altman, CEO of OpenAI, stated that this decision aims to ensure an optimal experience for all users.

A significant environmental impact 🌱💧

While image creation via ChatGPT with GPT-4o offers fascinating possibilities, its ecological footprint must be considered. Each image generation draws on energy-intensive data centers, contributing to CO₂ emissions. To put it into perspective, producing 1,000 images with this model equals the emissions of a 6.5 km drive in a gasoline-powered car.

Additionally, these data centers use vast amounts of water to cool servers. It is estimated that an average ChatGPT session, consisting of 10 to 50 responses, can consume up to half a liter of fresh water. This impact raises concerns about the future of AI and sustainable development.

It is crucial to use these technologies responsibly, being aware of their environmental costs. OpenAI and other industry players are encouraged to develop more sustainable solutions to mitigate these effects.

A revolution, but one to watch 🌟

The integration of GPT-4o into ChatGPT marks a turning point in AI-assisted image creation. By making image generation more accessible and precise, OpenAI opens new horizons for creators, designers, and anyone looking to enhance their visual content. But in the era of responsible digital development, these advances also raise questions about the ecological footprint of AI models. It remains to be seen how OpenAI and other tech giants will tackle this challenge.

Have you used ChatGPT before, or do you use it regularly? What do you think of this new feature? Give us your opinion in the comments section.