NVIDIA recently hosted the GeForce RTX AI PC Tour in Bangalore and showed off a bunch of RTX technologies, some prototypes and some fully functional, aimed at gaming, content creation, development, and even academics. We got to visit the event and get a first-hand look at these emerging AI technologies with engaging hands-on demos, led by John Gillooly, Technical Product Marketing Manager, Asia Pacific South at NVIDIA.
NVIDIA ACE (Avatar Cloud Engine) and Digital Human Technology
NVIDIA’s tour highlighted several key AI innovations, the first of which was the NVIDIA ACE (Avatar Cloud Engine) technology, backed up by the Digital Human Technology which uses NeMotron-4, a small language model with 4 billion parameters, bringing game characters to life with advanced conversational abilities and AI-powered vision.
Demonstrated through upcoming titles like Amazing Seasun Game’s Mecha Break and Perfect World’s Legends, ACE allows in-game characters to engage in intelligent conversations with players as well as other lifelike NPCs in games and in the case of Legends, even recognize real-world objects via the computer’s camera powered by ChatGPT-4o.
In our time with the demo, we got to interact with the character “Yun Li” using a microphone in real-time, and tried to test its limits by asking various impromptu questions. Utilizing a form of “guard-rail” system as explained by John, the NPC can stay on topic, remember older conversations, and even hold grudges if need be. When needed, they can talk about random topics such as the weather or even life, but ask her about the quest and she’ll immediately brief you on all the details as naturally as possible.
We found this to be the most promising of the demos, with possibilities of random NPCs being given their own personalities, backstories, and even memories that evolve over time as the player progresses through the game. Best be careful when trying to loot these NPCs by placing a bucket over their head next time.
The Multitude of AI Innovations for Gaming and Creativity
RTX Remix
RTX Remix, an AI-driven solution for modding older video games allowing for the introduction of ray-tracing features and other enhancements such as real-time upscaling and rendering was also showcased. This holds much promise too as numerous older classics can be given the RTX treatment allowing for a newer generation of gamers to experience them with enhanced visual fidelity, the way our nostalgia remembers it.
DLSS 3.5 with Ray Reconstruction
NVIDIA also demonstrated how DLSS 3.5 with Ray Reconstruction elevates visuals by showcasing the partner game Star Wars: Outlaws, Massive Entertainment’s latest outing under Ubisoft utilizing the powerful Snowdrop Engine, providing stunning, photorealistic graphics even on mid-range GPUs. Combined with NVIDIA Reflex, which reduces system latency, players can experience smoother, more responsive gameplay.
Omniverse Audio2Face
Beyond gaming, NVIDIA showcased Omniverse Audio2Face, a tool that transforms audio into realistic facial animations and lip-sync with the click of a few buttons. Intuitive Character retargeting allows the user to connect and animate their own characters.
In our time with the demo, we got to see John’s audio transformed into a facial expression in real-time on the screen, where the parameters for the expression were also fine-tuned. This holds incredible promise, especially for indie developers who can animate their own characters with the click of a few buttons and avoid the likes of certain notorious games for having dead expressions on release. Once satisfied we left the station with the animation paused midway, the demo character staring longingly into the distance as the rest of the event continued.
ComfyUI
Next up we got to try out one of the most popular Stable Diffusion applications with 10x faster image generation compared to online alternatives. It’s worth noting that most of these demos can run completely offline on the NVIDIA GPU, not requiring any form of online connection which is only optional.
This demo showcased the flexibility of having a selfie taken on the spot using a webcam being turned into superhero representations of ourselves in a matter of seconds, thanks to the existence of TensorRT and the enablement of open-source AI tools. This is another useful segment for creators or advertisers who’d like to create ideas or even final products for say thumbnails in social media.
The workflow being incredibly flexible is a huge bonus, and we came out of this station satisfied with the six-pack abs we’d been endowed with.
ChatRTX
A standout demo, ChatRTX, leverages retrieval-augmented generation (RAG) and TensorRT-LLM to create custom AI chatbots, completely personalized to your data. It can be docs, notes, images, or anything else. In our time with the demo, we got to make it retrieve a section from a particular document by just typing a vague description of what we wanted. Running locally on RTX PCs, ChatRTX allows users to query chatbots securely without the need for an internet connection. The possibilities for this are also endless. Imagine having your own personalized AI assistant chastize you for not cleaning up your inbox, with no worries about privacy leaks due to the offline nature.
RTX-Powered Laptops: Performance on the Go
The tour also featured NVIDIA’s RTX 40 Series laptops, which cater to gamers, creators, and professionals alike. These laptops, equipped with RTX GPUs, deliver double the AI performance of competing devices like the Apple MacBook Pro M3 Max. Featured models include the ASUS TUF A15, HP Victus, Lenovo LOQ, Acer Aspire 7, MSI Crosshair 16HX, and Dell G15. NVIDIA and its partners are set to launch a comprehensive line-up of such GeForce RTX 40 Series laptops, keep a lookout for these.
At the core of NVIDIA’s RTX offerings are Tensor Cores, AI accelerators that power every GeForce RTX GPU, enabling advanced AI workloads in gaming, content creation, and STEM applications. Delivering between 200 and 1,300 trillion operations per second (TOPs), GeForce RTX systems outperform traditional neural processing units (NPUs), which max out at just 45 TOPs.
With over 100 million GeForce RTX and NVIDIA RTX GPUs in use globally, NVIDIA ensures that users can leverage AI for more than 500 AI-enabled applications and 370 AI-enhanced games. These AI systems support leading large language models (LLMs), including Google Gemini, Llama 2, and OpenAI Whisper, making them versatile across industries.
NVIDIA Broadcast and NVIDIA Canvas
We got to close off the event by trying out NVIDIA Broadcast and NVIDIA Canvas. NVIDIA Broadcast is something I’ve personally used for a long time now during my gaming sessions with friends, and it’s pretty efficient at isolating my vocals from the keystrokes of my mechanical keyboard or other disturbances elsewhere. It’s also capable of doing enhancements to videos by rendering different backgrounds, which is incredibly useful if you’re sitting for a business meeting with your laundry behind you.
And NVIDIA Canvas was pretty intriguing in that it turned simple brushstrokes into realistic landscape images. This could be useful to visualize ideas for concept exploration and even create simple backgrounds for your next DnD session quickly perhaps.
Following the advancements in these AI technologies has been fascinating in the speed it’s been developing and it’ll be interesting to see how consumers utilize these now readily available technologies to further the feedback loop of creativity, innovation, and development. We’ll keep reporting on such tech developments and hope for similar open demonstrations and demos from companies like NVIDIA and others in competing technologies, so stay tuned.
