The new AI created by Google makes incredible and realistic images from the description of an idea. Beats DALL-E 2

'': when MICROSOFT launched ANTI-GOOGLE ads

Text-to-image broadcast model

IMAGE is, as its creators explain, "a model of diffusion of text to image" in which the results are realistic, as we can see in the results. IMAGE is based on artificial intelligence of linguistic models for text comprehension and relies on diffusion models for the generation of high-fidelity images.
To evaluate text-to-image models in greater depth, this website has also created DrawBench. With DrawBench, IMAGE is compared with recent methods, such as VQ-GAN+CLIP or with DALL-E 2.

Closed to the public

Jeff Dean, senior artificial intelligence researcher at Google AI, has shared numerous images on his Twitter profile.
AI can unlock joint human/computer creativity! Image is one direction we are pursuing:
"A high contrast portrait of a very happy fuzzy panda dressed as a chef in a high end kitchen making dough. There is a painting of flowers on the wall behind him."

— Jeff Dean (@JeffDean) May 24, 2022

According to Google researchers, it is currently only available as a scientific document. For ethical reasons, it is unlikely that this will change in the near future, but they do not rule it out in the future.

Via | Carlos Santana

Related Posts

Leave a Reply

%d bloggers like this: