At the end of March, OpenAI released an updated version of ChatGPT’s image generation tool with „4o“. This is yet another proof of how powerful artificial intelligence (AI) technology is. Moreover, it demonstrates the rapid pace at which today’s AI tools are advancing, offering even more sophisticated capabilities.
The New Version Brings Enhanced Features
The new ChatGPT image generation version comes with improved capabilities. Latest update stands out for its ability to accurately recreate requested prompts, drawing from a vast knowledge base and identifying all details within a conversation.
„4o“ has replaced the previous DALL·E AI model, though DALL·E will not be removed, and users will still have access to it. The updated version is available across various plans: Plus, Pro, Team, and Free (although access is currently limited due to high demand).
What’s New in „4o“ Version?
When introducing the updated image generation tool, OpenAI highlighted the following aspects:
- Improved contextual interaction. ChatGPT’s image generator was trained using online images along with text. During the training process, significant emphasis was placed on understanding visual content to ensure images are not only visually appealing but also highly accurate and realistic.
- Enhanced text representation. Previous models struggled with properly controlling and rendering text within generated images. Now, the system can accurately and grammatically correctly insert words into visuals.
- Multi-Level interaction. The „4o“ now offers enhanced interaction capabilities. Within the same conversation window, the tool will retain previous context, reducing the need to restart the entire process for each new request.
- Greater detail. OpenAI emphasizes that image generation will be extremely precise. It is reported that the model can effectively process 10–20 different objects in a single request.
- Additional input options. Users will now be able to provide not only textual prompts but also additional images, which the system can use as references for imitation.
The company acknowledges that, as this is a new update, some inaccuracies were observed during testing, which will be addressed over time:
- Hallucination risk and editing inaccuracies. There is still a risk of hallucinations, particularly when the image generation prompt is imprecise. Additionally, generated images may sometimes be incorrectly cropped.
- Limited object capacity. As mentioned, 4o can handle up to 20 objects effectively, but a higher number of elements may lead to less accurate results.
- Language limitations. While text grammar improvements are underway, the model may still struggle with certain languages it has not been trained on extensively.
- Multi-Level interaction challenges. The system does not always maintain full consistency during multi-step interactions, which could introduce challenges in the editing process.
The New Version Has Gained Massive Popularity
The ChatGPT image generation update has gained significant traction in just a few days. Users rushed to test the incredible capabilities of this AI model. Viral images began spreading across social media, showcasing photos transformed into Studio Ghibli-style illustrations.
A Studio Ghibli-style image imitating one of the popular social media memes, where a girl smiles in front of a burning building. Source: X
Due to this massive surge in activity, OpenAI temporarily limited access for free account users. Many were impressed by the tool’s ability to generate highly realistic images, while others were excited about its improved text rendering capabilities. Below are some examples shared by users on social media platform X.
A user requested „4o“ to generate an image similar to Johannes Vermeer’s painting „The Girl with a Pearl Earring“. Source: X
The user’s text prompt, along with a sample template, was transformed into real photographs. Source: X
Final Thoughts
We are living in a true technological revolution. The case of ChatGPT’s „4o“ illustrates how rapidly AI is evolving. In the future, we can likely expect even greater breakthroughs and unimaginable capabilities that will inevitably transform both our personal lives and careers.
If you are interested in this topic, we suggest you check our articles:
