OpenAI Launches Its Most Advanced Image Generation Tool Integrated Into GPT-4o

Main Image
  • Like
  • Comment
  • Share
TL; DR
  • The new image generation capability is built into GPT-4o, i.e., it is not a separate tool.
  • Among the improved capabilities of the GPT-4o image generation tool include text rendering.
  • The GPT-4o has gotten better at understanding natural language and refining images in multiple steps.
  • All generated images come with C2PA metadata, which helps identify an image generated using GPT-4o.

On March 25, 2025, OpenAI announced its most advanced image generation tool built into GPT-4o. While language models have been able to generate images for a while, OpenAI claims that its latest text-to-image model offers improved text rendering. Further, the company says the tool is better at following instructions than ever.

The New Image Generation Tool Is Built Into GPT-4o

OpenAI Launches Its Most Advanced Image Generation Tool Integrated Into GPT-4o

First and foremost, the new image generation capability is built into GPT-4o, i.e., it is not a separate tool. As seen in the samples shared by OpenAI, the language model seems to have gotten much better at following text-based instructions, not just those with a few words but ones with paragraphs of descriptions.

“GPT-40 image generation excels at accurately rendering text, precisely following prompts, and leveraging 40’s inherent knowledge base and chat context — including transforming uploaded images or using them as visual inspiration,” mentions the official press release.

Also Read: Sony Refreshes Its Affordable TWS Earphones With WF-C710N, Could Debut In India By June 2025

The Tool Offers Improved Text Rendering And Natural Language Understanding

Among the improved capabilities of the GPT-4o image generation tool include text rendering. The language model can blend precise symbols with imagery. For instance, if you ask GPT-4o to generate an image of a city’s signboard with the instructions given in it, it will generate a life-like image. Similarly, you can ask the tool to create an image of a restaurant’s menu by entering the description of the dishes.

The GPT-4o has gotten better at understanding natural language and refining images in multiple steps. Suppose the language model has generated an image; you can ask it to edit the image with the required changes, and the character or subject in the image will maintain its appearance across multiple iterations.

OpenAI’s image generation tool has gotten better at following instructions, especially when 10-20 objects are in the image. Last but not least, the tool learns from user-uploaded images and integrates its learning and the context into image generation. In other words, you can ask the tool to create a diagram of a complex scientific phenomenon, and it will do so with ease.

Also Read: realme P3 5G Goes On Sale With Rs. 2,000 Bank Discounts: Check Variants And Prices Here

GPT-4o’s Image Generation Tool Falls Short In The Following Areas

OpenAI Launches Its Most Advanced Image Generation Tool Integrated Into GPT-4o

Along with its pros, OpenAI has also elaborated on the cons of the GPT-4o’s image generation. The issues include unnecessary cropping, hallucination (a phenomenon when language models start making up information), and struggles with rendering more than 20 objects accurately.

OpenAI Has Put In Several Safety Systems In Place As Well

The company is also taking care of the safety concerns associated with generating realistic images. For instance, all generated images come with C2PA metadata, which helps identify an image generated using GPT-4o. OpenAI has also developed an internal search tool that helps determine whether an image was generated using its model.

The company has also placed barriers to prevent misuse related to child sexual abuse and sexual deepfakes (including robust safeguards around graphic violence and nudity).

With all the safety systems in place, GPT-4o image generation is already available to Plus, Pro, Team, and Free users as the default image generator in ChatGPT. Enterprise and education users will soon gain access to the tool. OpenAI’s older image generator, DALL-E, is still available via DALL-E GPT.

Also Read: realme P3 5G Goes On Sale With Rs. 2,000 Bank Discounts: Check Variants And Prices Here

You can follow Smartprix on TwitterFacebookInstagram, and Google News. Visit smartprix.com for the latest tech and auto newsreviews, and guides.

Shikhar MehrotraShikhar Mehrotra
A tech enthusiast at heart, Shikhar Mehrotra has been writing news since college for an undergraduate degree in Journalism and Mass Communication. Over the last four years, he has worked with several national and international publications, including Republic World, and ScreenRant, writing news, how-to explainers, smartphone comparisons, reviews, and list-type articles. When he is not working, Shikhar likes to click pictures, make videos for his YouTube channel, and watch the American sitcom Friends.

Related Articles

ImageiPhone 17 Nears Production As Apple Reportedly Completes Engineering Validation Testing (EVT)

The iPhone 17 lineup is expected to arrive sometime in September 2025. Months ahead of its official announcement, a DigiTimes report claims that Apple has completed engineering validation testing (EVT) for an iPhone 17 model. For those catching up, EVT refers to a phase in product development where prototypes are evaluated for hardware functionality. Also …

ImageThis Is How I Create My Ghibli-Style Portraits For Free (And You Can Do It Too)

Most recently, OpenAI announced the rollout of its most advanced image generator (as part of GPT-4o). In no time, the internet put the tool to work for generating Ghibli-style portraits of memes, iconic movie scenes, popular action heroes, and, above all, personal portraits. What Are Ghibli-Style Portraits Anyway, And Why Is The Internet Going Crazy …

ImageOpenAI Releases ChatGPT Pro, A $200 Monthly Subscription Model With Maxed Out Capabilities

TL;DR After GPT-4o (which was available for all users), OpenAI has now launched ChatGPT Pro, its most expensive consumer-grade subscription to date. Made for professionals like researchers and engineers, the ChatGPT Pro subscription lets users access the company’s most powerful computational models, which include o1, o1-mini, Advanced Voice, and o1 Pro mode.ChatGPT Pro Costs …

ImageOpenAI Releases Deepfake Classification Tool For Images Generated With DALL-E 3

Amid rising cases of deepfakes surfacing on the internet, OpenAI, the company that made ChatGPT, has released a new tool to help curb the spread of fake photos and videos. On Tuesday, the company announced its image detection tool, which helps determine whether an image has been generated using the DALL-E 3 text-to-image tool. Also Read: …

ImageElon Musk Slams Microsoft for using Scarlett Johansson’s AI-Generated Voice in GPT-4o

Artificial Intelligence is taking over the tech world in both good as well as bad ways. While AI proves to be beneficial in some situations, there are times when it has been questioned for violating the privacy of individuals. Recently, Hollywood Actress Scarlett Johansson accused OpenAI of using her voice for the new ChatGPT-4o model …

Discuss

Be the first to leave a comment.