UPDATE: OpenAI has halted access to the image generator for free users amid high demand. “Images in chatgpt are wayyyy more popular than we expected (and we had pretty high expectations),” CEO Sam Altman tweeted this afternoon. “Rollout to our free tier is unfortunately going to be delayed for awhile.” (It had to do the same thing with the Sora video generator in December.)
Original Story:
OpenAI has added AI image generation capabilities to ChatGPT. Users can now select the GPT-4o model, provide prompts, and get desired images within the regular ChatGPT window.
Previously, ChatGPT was dependent on OpenAI’s DALL-E model for images. Now, it uses the 4o model’s native multimodal capabilities to provide “precise, accurate, photorealistic outputs.”
OpenAI touts GPT‑4o’s skill for “accurately rendering text, precisely following prompts, and leveraging 4o’s inherent knowledge base and chat context—including transforming uploaded images or using them as visual inspiration.” Translation: Expect fewer weird results.
This was achieved by training the models on “the joint distribution of online images and text, learning not just how images relate to language, but how they relate to each other,” OpenAI says.
OpenAI’s demos for images containing text (Credit: OpenAI)
GPT-4o can also handle more objects within an image than usual. While other chatbots can generate up to eight objects for an image, GPT-4o can produce up to 20, according to OpenAI.
It can also edit and improve user-uploaded images. In a demo video, an OpenAI researcher is seen uploading a hand-drawn sketch for a comic book page and getting a full-colored digital version delivered by ChatGPT.
Still, OpenAI warns, “Our model isn’t perfect. We’re aware of multiple limitations at the moment, which we will work to address through model improvements after the initial launch.”
OpenAI will embed each output with C2PA metadata. This will allow AI image detectors to identify images generated by GPT-4o accurately. Additionally, ChatGPT will reject requests for child sexual abuse materials (CSAM) and sexual deepfakes. “When images of real people are in context, we have heightened restrictions regarding what kind of imagery can be created, with particularly robust safeguards around nudity and graphic violence,” OpenAI says.
Recommended by Our Editors
In an addendum added later, OpenAI said it won’t block GPT-4o from generating images of adult public figures, but those “who wish for their depiction not to be generated can opt out.”
At launch, ChatGPT’s native image generation is available for all Plus, Pro, Team, and Free users, with support for Enterprise and Edu customers coming soon. The feature is also available on OpenAI’s video-generation tool, Sora.
OpenAI hasn’t announced a daily limit for free users but tells The Verge that it will mirror DALL-E, which limits users to three free images per day. However, these numbers “may change over time based on demand,” a spokesperson adds.
None of this means DALL-E is going away. “For those who hold a special place in their hearts for DALL-E, it can still be accessed through a dedicated DALL-E GPT,” OpenAI says.
Get Our Best Stories!
This newsletter may contain advertising, deals, or affiliate links.
By clicking the button, you confirm you are 16+ and agree to our
Terms of Use and
Privacy Policy.
You may unsubscribe from the newsletters at any time.
About Jibin Joseph
Contributor
