OpenAI, the company behind DALL-E and ChatGPT, has just released to the public an artificial intelligence tool that generates 3D models from text and renders. The program, called Point-E, needs to run locally, but it doesn’t require state-of-the-art hardware to do its job.
According to an OpenAI article, the tool is capable of generating point clouds that represent a 3D shape using an Nvidia V100, that is, a graphics card with Volta architecture that was launched on the market just over five years ago. Point-E needs a text (prompt) or a render as a starting point.
More tools for 3D rendering
OpenAI is not the only company that is using AI to create these types of tools. For example, Google has DreamFusion, but the program can take several hours to create 3D models. The solution from the company led by Sam Altman promises to do the same job, in just a few minutes.
The text-to-image generation model works similar to DALL-E, Midjourney, or Stable Diffusion. If we ask you for an image of a chair shaped avocado, the model will generate a 3D rendering of a point cloud of our request. We can also ask you to generate a point cloud from a model that has been generated with another tool.
As we can see, 3D point clouds are far from being perfectly designed elements, but that has an explanation. These are frequently used by designers for 3D modeling tasks for a wide variety of tasks. From manufacturing objects to creating complex animations for the film and video game industries.
OpenAI assures that it will continue working to improve the capabilities of Point-E. The good news is that, as we said, the program in its current phase can be downloaded to test locally. Although, yes, this possibility is limited to users with certain technical knowledge.
Downloading the code to the computer will not be enough. Additional software, such as a Jupyter notebook, will need to be downloaded to run the code in the web browser. The project’s GitHub profilein addition to providing the OpenAI resources, provides information to make this new AI tool work.
Many of the possibilities that AI has presented were unimaginable just a few years ago. In a short time, imaging tools and chatbots have evolved by leaps and bounds, and they seem to be here to stay. It will be interesting to see the progress of these new 3D modeling tools.
Images: Ashley Whitlatch
In Xataka: There is an AI to cancel the noise when eating chips. We have tested it to see if it works
