what is surprising is that it can even create videos or develop images

OpenAI’s DALL-E isn’t the only artificial intelligence capable of generating images from a brief textual description. A few weeks ago, Google also presented “Image”, an AI alternative from the company founded by Elon Musk (among others) which, according to the Mountain View firm itself, is capable of creating designs much more realistic and of better quality. Today, Microsoft has joined the competition. He does it with NUWA Infinity an AI that is not only capable of producing images from texts, but also to convert a static drawing into a video.

Microsoft describes NUWA as “a multimodal generative model designed to generate high-quality images and video from given text, image, or video.” Its operation is therefore not very different from what DALL-E or even Image (Google) can do. However, it has a number of advantages over the two models of artificial intelligence. It is the only AI capable of generating a video from an image created by a text description. AI, moreover, can also generate a video directly from a description.

Compared with DALL-E, Imagen and Parti, NUWA-Infinity can generate high resolution images of arbitrary size and also support long video generation.

NUWA, Microsoft’s AI can also extend any type of image

NUWA, Microsoft’s AI that generates images and videos from a textual description, is also capable of… “stretch” any image and create a larger, higher resolution image. Artificial intelligence, in particular, detects the information contained in the original photograph and, depending on its parameters, generates another much more complete one. NUWA, for example, can extend Vincent van Gogh’s “Starry Night”. It does so, moreover, with a detail identical to that presented in the original design and a very precise continuation.

At the moment, Microsoft has not given more details about NUWA, beyond a few examples that show the potential of this AI and how it is able to convert text to image, image to video or text to video. , as well as the ability to extend any design. It is certainly an interesting alternative to DALL-E and Imagen, although these two algorithms also have their advantages.

Picture, for example, generates much more realistic drawingsalthough it is not yet available to users. DALL-E, on the other hand, offers less realistic images, but is more accessible to users, as it is available through a public beta albeit with limited access.

Leave a Comment