Intel has just launched LDM3D, a new generative artificial intelligence capable of creating 360° 3D virtual environments from a written request. The AI is available under a free license.
This will also interest you
[EN VIDÉO] An artificial intelligence that adjusts the angle of taking the photo Interactive manipulation based on points of the generative image collector. © DraGAN Project
Intel Labs has just launched a new kind of generative artificial intelligence. The Ais capable of generating an image from a sentence, such as Midjourney or Dall-E, are starting to be quite well known. However, these images are always in two dimensions. In collaboration with Blockade Labs, IntelIntel Labs has announced its LDM3D model (Latent Diffusion Model for 3D) for the generation of 360° and three-dimensional images.
LDM3D is based on a specialized version of Stable Diffusion version 1.4. The AI generates an image of a 360-degree environment in two dimensions, as well as a depth map (depth map) which contains information about the relative distance of each pixelpixel. This model can generate in a few moments a peaceful forest, an urban landscape or a futuristic sci-fi world.
Some examples of environments generated by LDM3D. Drag the images to explore them in 360 degrees. © Intel Labs, Blockade Labs
A model published under a free license
The two teams then created DepthFusion, a system that uses the visual programming language TouchDesigner to create “immersive and interactive digital digital experiences”. In other words, it combines the image and the depth map to create a 3D and 360-degree virtual environment.
The images generated by LDM3D could revolutionize the creation of content for virtual reality and the metaverse, as well as video games, video games and even architecture. The model is available as free software on the Hugging Face website.