OpenAI makes its new image generator Dall-E 3 generally available via ChatGPT. The AI ​​can produce more beautiful and detailed images than its predecessor.

OpenAI makes Dall-E 3 accessible to a wider audience. The AI ​​model appears in ChatGPT, but currently only paying users can work with it. You must have a ChatGPT Plus or ChatGPT Enterprise subscription.

Dall-E 3 is an image generator based on generative AI. OpenAI impressed the world with the initial launch of Dall-E, even before ChatGPT really made a splash a few months later. Like its predecessors, Dall-E 3 understands descriptions of images and generates images from them. OpenAI launched the image generator last month.

Improvements and Limitations

The resolution of these images remains limited to 1,024 x 1,024 pixels in this third version of the AI. Other things have improved. In this way, Dall-E 3 understands more complex questions and the images have more details. Dall-E 2 certainly had some problems with human or animal body parts. This third version can handle that much better.

All images you generate with Dall-E 3 via ChatGPT are intended for public use. So you can do whatever you want with it. OpenAI also claims to have built in better protection against malicious content. When it comes to copyright protection, OpenAI goes one step further by making it impossible to ask for an image in the style of a living artist.

Work

To get started with Dall-E 3, select the image generator at the top when starting a new chat conversation.

Then you can let your imagination run wild. Be very specific: One of the qualities of Dall-E 3 is precisely its ability to process detailed and concrete prompts. So don’t hesitate to describe your wishes in detail. ChatGPT understands Dutch, but the prompts are translated into English for Dall-E.

During our testing we see significant improvements, although the images are still not perfect. Feet and kicks are a challenge and we wonder where the lady’s second leg on the left is. Still, faces, arms, and even hands are much more realistic than in previous versions. For this print we asked for an image of a person with the face of a cat on the stairs of the London Underground.

In this second printing we asked specifically about a pantomime performance by a cat person in front of an applauding audience on a sunny street in Paris. Here too, one or two people sprained their hands a bit, but everything looks much more realistic. When you click on an image, you immediately see a description of Dall-E 3 itself, which serves as a kind of justification for the work and its relation to your request.

Finally we decided on the atmosphere. Once again a cat takes center stage. This time she had to breathe fire on the roof of a building in New York while it was raining at night. Pay attention to details like the reflection of the flame on the cat and the building.

Dall-E 3 is just available and we are not the only ones who wanted to try the model. So it will take a while for the image generator to convert you into an expression immediately. During our tests we had to wait at least half a minute. Also, it is not always possible for us to download an image immediately using the button above.