DALL-E 2 is an artificial intelligence tool designed by OpenAI that can turn simple words into images. How does it work ? How to use it ? We tell you everything.
Artificial intelligence has made tremendous progress in recent years. This allows him to express himself in many areas, and this, in an already very promising way. OpenAI works on a wide variety of projects. Recently, the company made news with its algorithm DALL-E 2, which makes it possible to generate very complex images from simple descriptive texts. After reading this article, you will know everything there is to know about the subject.
What is DALL-E 2?
DALL-E 2 is a product of OpenAI, a startup supported in particular by the American giant Microsoft. It is an artificial intelligence model that has the particularity of being able to convert purely textual descriptions into images. AI models like this come from the Google Imagen software. The software developed these from collections of words and images on the Internet as well as photos posted on Flickr.
For an overview of how it works and sample images created by DALL-E 2, watch the video below:
If we had to describe DALL-E 2 in a few words, we could say that it is a “new AI system that can create realistic images and art from a description in language natural”. Simply insert words, any textual description of the image you “want” and DALL-E will convert the text to an image using its AI models.
Who can use this OpenAI AI?
Currently, OpenAI, like Google, does not offer this technology to the general public. Only a restricted community, via an invitation system, can take advantage of this truly impressive service. To be on the waiting list, the interested user must indicate whether they are a developer, professional artist, university researcher, journalist or online creator.
The purpose of this waiting list is multiple. First, it’s about limiting the influx of requests, and the resources used, by extension. When there are too many usage requests, a dialog box appears saying “Too much traffic, please try again”. But it also avoids putting this powerful technology in all hands. Some might use it for the wrong reasons.
Anyway, according to Joan Jang of OpenAI, “we are working hard to accelerate access, but it will probably take some time to reach everyone ; As of June 15, we have invited 10,217 people to try DALL-E.”
DALL-E 2 offers simple interfaces. This is the whole objective of the product, ultimately. A text area, a button to launch the generation and an area to display the created image. The team also added watermarks to the lower right corner of the images to indicate the source for DALL-E 2 and Imagen. This is to avoid potential damage. Moreover, a number of filters have been put in place to avoid violent content, nudity, conspiracies and other personalities. “There is a continuous process of improving the security of these systems,” says OpenAI researcher Prafulla Dhariwal.
Like artificial intelligence more generally, this type of product is very innovative. The quality of the results offered is therefore very variable. So don’t be surprised to see images that are quite different from what you imagined.
If you want to know more, go to the official DALL-E 2 website and to be able to test the tool, you will need to fill out the form on this page.