NVIDIA now wants to create 3D models just by describing them

While “Midjourney” for smartphones officially lands on the App Store, AI for image creation is now a reality with many fierce competitors. while, nvidia I was able to produce a tool that can also generate polygonal models.

he is called Magic3D It is an AI-based tool, for which the academic publication by Lin et al. is also available.

Its goal is to go and attack a market that it currently dominates Fusion dream, the result of research and development by Google. The main difference between the two lies in the method of propagation.

The NeRF model is used in DreamFusion Traditional, while Magic3D’s approach involves two basic steps, interpreting the input with low fidelity model generation, and then properly optimizing and upgrading it to make it usable.

According to the paper’s authors, Magic3D is twice as fast from Mountain View competitor. In addition, always by entering the text it will be possible to create specific grids to be applied to the model to produce the most reliable rendering possible.

An interesting project that adds to what has already been produced with publishing tools. Example? Our last experience where we asked Midjourney to finish God of War Ragnarok!

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *