There is an increase in the artificial intelligence projects which converts the terms that you have entered in the system to visuals. As MidJourney became one of the most popular examples, Google is also developing its own illustration technology.
The project had started last year with the name of Google Dream Fields, and it is based on an artificial intelligence-based 3D image rendering technology. This year, the name was changed as DreamFusion. The difference is that it is based on Google Imagen modelling technology rather than OpenAI CLIP modelling technology.
The Google DreamFusion technique uses the Neural Radiance Field (NeRF) artificial intelligence network to create complex 3D scenes using 2D data packets. The engine, which first takes the appearance as material color, then processes under different lighting scenarios.
The advantage of the Google DreamFusion technic is there is no need to train the artificial intelligence. Also, it is possible to create a new scene with these images. The project is available to use on GitHub.