OpenAI presented the third version of its artificial intelligence DAll-E 3 to generate images from text, which comes with several improvements and new features. Among these it stands out that DALL-E 3 is now natively integrated into ChatGPT, allowing users to create better prompts, requests, or instructions given to an AI model to generate a response, directly from the chatbot. Here is the demo video of OpenAI DALL-E 3.
According to Sam Altman‘s startup, this tool has evolved considerably compared to its predecessor when it comes to interpreting user requests. In fact, when producing an image you can ask the chat system to generate it or create a longer and more detailed indication, which will allow DALL-E 3 to interpret it better and offer more precise results.
In addition, it will also allow users to refine a creation as if they were asking an artist for changes, thanks to this integration with ChatGPT, which better understands how a scene should be composed and what the link is between the elements that are part of it. The same.
How the integration works
:quality(85)/cloudfront-us-east-1.images.arcpublishing.com/infobae/FZ3HWY4GDZG5TJ6WWNIWW3ZZW4.webp)
DALL-E uses what is known as a diffusion model in order to predict how to render an image for a given request. With sufficiently large amounts of data, you can produce complex, coherent, and aesthetically pleasing images.
The novelty of Dall-E 3 is that it eliminates some of the complexity necessary to refine the text sent to the program, known as “prompt engineering”, and allows users to introduce improvements through the conversational interface of Dall-E 3. ChatGPT.
“Modern text-to-image systems tend to ignore words or descriptions, forcing users to learn prompt engineering. DALL·E 3 represents an evolution in our ability to generate images that adhere exactly to the text you provide”.
OpenAI
For example, this tool produced the following image in response to the following request: “An illustration of a human heart made of translucent glass, standing on a pedestal in the middle of a stormy sea.” The sun’s rays pierce the clouds, illuminate the heart, and reveal a small universe inside. The quote “Find the universe within yourself” is etched in bold letters on the horizon.”
:quality(85)/cloudfront-us-east-1.images.arcpublishing.com/infobae/7PYJQEQO6RGJFCSA3G3ZR7TVMQ.webp)
Another notable result is this surreal image that was generated with the help of ChatGPT, from this suggestion: “A vast landscape made entirely of various meats spreads out before the viewer. Tender, succulent hills of roast beef, trees of chicken thighs, rivers of bacon, and rocks of ham create a surreal but appetizing scene. The sky is adorned with a pepperoni sun and salami clouds.”
:quality(85)/cloudfront-us-east-1.images.arcpublishing.com/infobae/HGHFQ4PRWRD5HPDKI2WB7WAT2Q.webp)
Normally, this would require a huge prompt engineering effort, however, with Dall-E 3, ChatGPT is in charge of creating that more sophisticated message.
DALL-E 3 promises better results and more security
DALL-E 3 promises better results when including text within drawings, as well as when dealing with parts of the human body that it previously failed to interpret correctly. For example, the hands.
In addition, it also implemented more robust security measures in order to mitigate bias and prevent the use of the tool to create potentially harmful content such as deepfakes, which can be a video, image, or generated audio that imitates the appearance and sound of a person. In fact, the tool will refuse to create images of public figures based on their names.
Open AI also reported that it will offer artists the ability to remove their works from future Dall-E 3 training, which will also prevent users from attempting to generate a work of art in the style of a well-known artist and has barriers to prevent users from generating pornographic or graphically violent works of art.
The new version of generative AI will arrive in October for ChatGPT Plus and Enterprise subscribers through its API.