Google has introducedImagen , anAI image generator .- Images is a text-to-image generator that uses a given description to generate images.
- Images can create drawings, CGI renders, oil paintings and more.
Google’s Imagen is similar to DALL-E 2 AI image generator, however, it is claimed to generate more detailed and accurate images. The generator can draw, paint, create oil paintings or even develop CGI renders, whatever you want it to do.
For example, if you enter “teddy bears swimming at the Olympics”, Imagen will generate the below image –
The below image is generated when you enter the text "a dragon fruit wearing a karate belt in the snow".
The image generator uses diffusion technologies which use a pure noise image and then slowly refine it until the AI can’t make it any better. Google has also used large language models that improve language understanding and convincing language models.
Imagen begins with taking the description and then generating a 64 x 64-pixel image. After two enhancements, the image is converted into a larger 1024 x 1024-pixel image. The AI creates new resolutions along with the smaller image.
You can find some of the images generated by Imagen here.
Do note that these are the images that Google has selected to showcase here. It may be possible that not all images come out this good and that the company has hand-picked the best images.
Google has claimed that Imagen consistently generates better images than DALLE-2, based on its DrawBench benchmark.
The
SEE ALSO:
Elon Musk says he doesn't use flight attendants in apparent denial of sexual-harassment allegation against him
A California bill could allow parents to sue social-media companies for up to $25,000 if their children become addicted to the platforms
Google’s new research on digital wellbeing needs your data to dig deeper into user behavior and reduce the risks of smartphone addiction