OpenAI’s ChatGPT may primarily be known for its advanced communication capabilities, but its latest GPT-4o image generation update is a reminder that it can do more than just talk a big game. The company has given users the ability to generate images for a while now, so the feature is nothing new, but OpenAI’s generative AI tools have leveled up significantly over recent months. If it’s been a while since you gave it a shot, you might be surprised by what it has to offer.

OpenAI’s latest AI art model has already sparked some controversy with its ability to duplicate the Studio Ghibli art style without any of the skill and effort that goes into creating the artwork. While Altman proudly embraced the copycat trend by changing his picture on Twitter/X to an image generated by the chatbot, we know Studio Ghibli’s Hayao Miyazaki has been staunchly against the capabilities of AI.

The OpenAI Studio Ghibli controversy is just another frontier for the conversation on plagiarism and theft of copyrighted works, but OpenAI’s generative AI tools are not going to take a step back just to accommodate it.

how to generate images with GPT-4o

Image: The GPT-4o image generator shows a candid paparazzi-style photo of Karl Marx generated with a very elaborate prompt.

GPT-4o Image Generation Update: What It Is and How to Use It

OpenAI’s GPT-4o is a multimodal large language model that showcases some of the latest AI capabilities presented by the company. The new flagship model was introduced on May 13, 2024, but OpenAI had been building on its capabilities ever since. Evolving from the text-based prompt system to accept a combination of text, audio, image, and video, eliminating the latency period between input and response to unbelievable degrees. 

This particular showcase of the GPT-4o image generation update on March 25, 2025, has shown us just how advanced the company’s image generation model is and how much users can expect to do with it in the future. 

How to Generate Images with GPT-4o

Users do not have to download any additional apps to experience the image-generation capabilities of GPT-4o. The AI image tools are available in ChatGPT through the website or the existing ChatGPT app directly. It can also be used in Sora or through a dedicated DALL·E GPT for those who still prefer the platform. Developers who want to generate images via the API will also soon be able to do so in the next few weeks.

Who gets access to OpenAI’s latest AI art model? Initially, it was stated that all Plus, Pro, Team, and Free users would be able to use the image generation capabilities of GPT-4o immediately, but in an update on March 27, Altman stated that free-tier users would have to wait a little longer for the rollout due to its unforeseen popularity. Enterprise and Edu members might have to wait a little longer, but they will be able to make the most of its capabilities as well. 

GPT-4o AI Art Features: What’s So Special About the Generative AI Tool

The release notes on the GPT-4o image generation update described what customers could expect with the tool’s upgrades. The post explained that GPT-4o’s advances have resulted in “image generation that is not only beautiful but useful,” showcasing how it can generate anything from realistic images to comic strips based on a few prompts.

The results are admirable, particularly in the example where OpenAI’s generative AI tools generate an artsy infographic, and the same image is successively used to generate a more realistic depiction of someone using the infographic in a notebook. What is most impressive and alarming about AI art is its capability of building on results with simple instructions.

Some of the other features and highlights showcased in the GPT-4o image generation update include: 

  • Consistency across images and context-aware interpretation of prompts
  • Ability to interpret simple prompts without requiring heavy detailing from the user. Simultaneously, it can also render more complex prompts into images
  • Ability to refine images through natural conversation and build on characters and contexts without losing the original design
  • GPT‑4o can handle up to 10-20 different objects, making it more detailed
  • Ability to take user prompts, including images, to utilize that information and context in its responses
  • Implements its world knowledge to interpret prompts to maintain consistency with realism, whether it’s in a recipe or interpretation of code to generate images
  • A variety of image-generation styles can be embraced, from photorealism to the art style at the heart of the OpenAI-Studio Ghibli controversy

Known Limitations of the OpenAI Generative AI Tools Presented by GPT-4o

As nice as it is to brag about advancements, AI tools are far from perfect. The GPT-4o image generation update acknowledges that the model still has some flaws, all of which it aims to address and improve post-launch. The image generator occasionally crops images and hallucinates information, especially in low-context prompts. 

When required to render more than 10-20 distinct concepts using its knowledge base, the AI can struggle. This is more obvious with the rendering of non-Latin languages, where incorrect information is still the norm. 

There are also a few issues that have been observed with asking the GPT-4o generator to edit specific aspects of the image, especially when it involves text. The AI can make incorrect adjustments or change parts of the image unprompted. OpenAI also acknowledges that the model struggles when asked to render detailed information at a very small size.

These issues are pretty standard for any image generation tool we have right now and reiterate the need for caution in their use. While they may be fun to use for generating infographics or How-to guides for your website or social media, you should ensure you check all the details thoroughly before you share it. The callous use of AI is one of the biggest challenges the industry faces right now.

OpenAI’s latest AI art model

Image: The detailed image generated of a photo taken with phone, of a glass whiteboard and a woman writing on it, is one of the most jaw-dropping AI images we’ve seen.

The GPT-4o Image Generation Update Addresses Safety and Security

OpenAI’s latest AI art model presents some fun use cases that its customer base has been experimenting with freely, but it’s also important to acknowledge the intent of use. With every advancement in AI, these tools only get more dangerous, perfectly primed for misuse. Google’s recent AI image generator upgrade was being misused to remove watermarks from images, which makes it much easier to steal art and intellectual property. 

While OpenAI can’t guarantee its users’ intentions, what it can do is try and make it safer to use. Generated images feature C2PA⁠ metadata to make it more identifiable as GPT-4o-generated material. The company has also built an internal search tool that should allow users to confirm that images were generated using its AI. 

Requests for abhorrent content such as child sexual abuse materials and sexual deepfakes are still prohibited, and restrictions have been heightened for images containing real people. These safety measures will require constant upgrading as malicious users are always on the lookout for workarounds and ways to exploit the system. 

OpenAI Studio Ghibli Controversy—Where Do You Draw the Line?

Since the GPT-4o image generation update, users have been enthusiastically using the tool to generate images and memes to their hearts’ content, enjoying the freedom it offers to those with zero art skills of their own. The ability to generate images that resemble Studio Ghibli art from movies like My Neighbor Totoro, Kiki’s Delivery Service, or Spirited Away has delighted users, who are recreating images of themselves or even means in this beloved art style.

While the images look adorable and the art style is addictive to see, there are conversations to be had about where we are headed as a society with short-cut technology like this, which builds on the hard work of other artists to master a skill that isn’t its own. The results are far from accurate, and neither are they particularly unique compared to the Snapchat filters we’ve seen before, but the ability to imitate a style is where the controversy lies.

Other AI image-generation models can also accomplish this, but OpenAI’s latest AI art model has better results, which is great for the user but not so much for the artists who spend years honing their craft and developing their own unique styles. These art styles are not necessarily protected under copyright law so the original artists may not be able to take any legal action, but it does show that with each advancement in AI, things tend to get murkier.

How Do We Feel About the AI Image Tools in ChatGPT?

For the most part, users have been enjoying using GPT-4o’s AI art features and image-generation capabilities to their full extent. We see these amine filter trends make the rounds every once in a while, so it’s likely that this hype of turning yourself into a Studio Ghibli character will die down soon enough, but it’s too soon to start seeing the downward decline. Art style imitation aside, the capabilities of this art generator are exceptional. 

Copyright infringement laws and other AI regulations are barely trying to keep up with the advancements of these tools the way they need to be, and we predict this will continue to be the case in the years to come. As such, the unrestricted and unmonitored growth of AI can only be checked by users while the CEOs battle it out in their power struggles to stay on top. 

OpenAI is never ashamed about “borrowing” data without permission, much like the controversy over imitating Scarlett Johansson’s voice or sharing paywalled content from the New York Times. Altman’s callous attitude towards the current situation of stealing a style and his ability to turn anything into a situation where he is the victim suggests that things are unlikely to change.

Users will have to exercise their discretion to determine how they want to use the AI image tools in ChatGPT or other apps that are promoting their own capabilities in 2025.

tagged in AI image tools in ChatGPT, GPT-4o AI art features, GPT-4o image generation update, how to generate images with GPT-4o, OpenAI generative AI tools, OpenAI Studio Ghibli controversy, OpenAI’s latest AI art model
Share: