I built real-time photorealistic avatars running entirely on Quest 3 hardware. It uses feed-forward Gaussian Splatting to hit >60fps on mobile VR chipsets and runs through WebXR, so users can join calls directly from their browser without downloading apps.
The surprising part: users say it feels more "real" than video calls despite being rendered. Something about proper depth and eye contact tricks your brain into feeling present with the other person.
The main technical challenge was balancing visual fidelity with Quest's performance constraints while keeping it accessible via web standards. Most systems either need desktop hardware or fall into the uncanny valley.
It’s not publicly available yet, but curious what HN thinks about the approach and potential applications in and beyond social VR.
randall•8mo ago
I'm into it, i have a quest pro (which i think is still higher res than Q3 but idk if it's powerful enough) and would be very interested in trying it out.
Not enough gaussian splatting research imo.
edwardahn9•8mo ago
It’s powerful enough for this application! Mostly because here, faces have way less vertices than the gaussian splat scenes that you see in research papers.
Totally agree that there’s not enough research, and honestly not enough people in VR making applications on top of it
edwardahn9•8mo ago
I built real-time photorealistic avatars running entirely on Quest 3 hardware. It uses feed-forward Gaussian Splatting to hit >60fps on mobile VR chipsets and runs through WebXR, so users can join calls directly from their browser without downloading apps.
The surprising part: users say it feels more "real" than video calls despite being rendered. Something about proper depth and eye contact tricks your brain into feeling present with the other person.
The main technical challenge was balancing visual fidelity with Quest's performance constraints while keeping it accessible via web standards. Most systems either need desktop hardware or fall into the uncanny valley.
It’s not publicly available yet, but curious what HN thinks about the approach and potential applications in and beyond social VR.
randall•8mo ago
Not enough gaussian splatting research imo.
edwardahn9•8mo ago
Totally agree that there’s not enough research, and honestly not enough people in VR making applications on top of it