Nvidia’s deep learning technologies continue to amaze and perplex us and we have seen how the company can utilize AI to automatically match vocal lines to 3D animated faces a few weeks ago.
This is the kind of fascinating technology that can enable people to produce wonderful things with ease, or, in the case of Nvidia’s latest announcement, possibly dreadful things with ease.
Nvidia’s GauGAN was originally seen launched a few years ago. It was shown that simple drawings may be transformed into photorealistic graphics with the push of a few buttons. It’s quite cool stuff, and it’s worth experimenting with. GuaGAN2 is now available, and it doesn’t even require your sketches to create incredibly realistic landscape photographs.
Instead, the deep learning algorithm can convert words into astonishingly intricate pictures. It has been trained on 10 million landscape photographs and appears to know what it is doing. The sample shows how the image adjusts to accommodate additional text as it is inserted, implying that modifications can be made quickly.
GuaGAN 2 works by combining the shape-to-image technology shown in GauGAN with text-to-image production. This means that these visuals can be created using a combination of drawings and words. Users can begin with a broad statement to get a sense of the overall picture and then go in and change the fine elements by hand.
If you want to experiment with some of this technology, you may do so on Nvidia’s interactive AI demos, however, we had some issues getting it to function. Alternatively, if you own an Nvidia RTX GPU, you can get Nvidia Canvas for free. It gives you a good flavor of painting with AI, and it’s cool to see it work in real-time with your designs.
In terms of gaming ramifications, essentially everything deep learning is likely to save developers time and people can create test scenes in a matter of seconds with the technology. It has a plethora of practical applications for anyone wishing to develop.