Project R2X: Nvidia’s new artificial intelligence assistant is a realistic avatar available on PC

Nvidia has presented Project R2X, a Artificial intelligence (AI) in the form of a realistic-looking PC avatar that has the capabilities to analyze and understand what is displayed on the desktop and, based on this, help in the use of applications, video conference calls and summarize documents, among others. functions.

The technology company has announced its new R2X project within the framework of the news announced at the CES 2025 technology fair, where it has indicated that it is an assistant designed to put a “face” to AI.

In this sense, Project R2X is an assistant that, instead of interacting like a ‘chatbot’, is shown in the form of an avatar and, therefore, interacts with users like a realistically represented animated character.

Among its capabilities, Nvidia has indicated that the avatar can see what is being displayed on the desktop and analyze said information to, around this, carry out different functions, as it has shared in a statement on its website.

For example, R2X can provide detailed information to the user, as well as help with the use of desktop applications or during video conference calls. Following this line, the assistant can also perform other tasks such as reading and summarizing documents, all through interaction with an animated character that acts realistically through the PC screen.

To achieve this look, the avatar is rendered using Nvidia RTX Neural Faces technology, a new generative AI algorithm that augments traditional rasterization – the process by which an image in vector graphic format is converted into a set of pixels – with “ fully generated pixels.”

Additionally, the avatar’s face is animated with a new model called Nvidia Audio2Face-3D, which improves the movement of the lips and tongue for a more natural look.

The technology company has also explained that, to carry out its tasks, R2X can connect to AI services in the cloud, such as the GPT4o model from Open AI or Grok from xAI. Additionally, it can also run on models built into Nvidia’s NIM microservices and AI Blueprints.

NIM microservices and new models Call Nemotron

This assistant, which has been presented as a preview, is based on new basic models developed by Nvidia, which run locally on NVIDIA RTX PCs and are designed to power “digital humans, content creation, productivity and development” , as he pointed out in a statement on his blog.

Specifically, the new models are accelerated by the new GeForce RTX 50 Series GPUs, and are offered integrated with Nvidia Inference Microservice (NIM) services. That is, the company’s set of inference microservices that brings together scalable AI-based software solutions to streamline the implementation of this technology in various use cases.

Within this framework, Nvidia has announced that it will launch a series of NIM microservices for RTX AI PCs from the main AI model developers, such as Black Forest Labs, Meta, Mistral and Stability AI. With this, it will cover use cases such as LLM, vision language models, image generation or PDF extraction, among other capabilities.

Following this line, it has also presented its new family of Llama Nemotron models which, composed of the Nano, Super and Ultra versions, have been designed as open LLM models that provide “high precision in a wide range of AI tasks.”

Specifically, the technology company has referred to the Nano version of the model, which will be offered as a NIM microservice for PC RTX AI, and which will allow tasks such as following instructions, coding and mathematics to be carried out.

Taking all this into account, the company has detailed that Project R2X is a demonstration of how RTX users and developers can use NIM microservices to create Artificial Intelligence (AI) agents and assistants.

Likewise, Nvidia has indicated that to carry out their own projects, developers will be able to download, configure and run these NIM microservices starting in February on Windows 11 PCs through the Windows Subsystem for Linux (WSL). Likewise, they will be compatible with devices that integrate the GeForce RTX 50 series, GeForce RTX 4090 and 4080, and NVIDIA RTX 6000 and 5000 GPUs.