VeyoLabs Cloud: Real-Time Raytracing on Next-Gen GPUs
How we achieved 24fps playback on 4K AI assets at zero latency — and what this means for real-time preview and live production workflows.
VeyoLabs Cloud: Real-Time Raytracing on Next-Gen GPU Infrastructure
Real-time raytracing has been commercially available in gaming since 2018. In professional content production, it has remained aspirational — the hardware cost of running path-traced raytracing at 4K and 24fps in real time was prohibitive outside of dedicated broadcast facilities.
VeyoLabs Cloud now delivers it as a standard feature of the production environment. Here is the technical architecture behind it.
The Latency Problem in AI Video Preview
AI-generated video has a persistent UX problem: you generate a frame or sequence, then wait. The generation completes, you review it, and decide whether to iterate. In a production session with many iterations, this wait time compounds into hours of dead time.
Real-time raytracing in the preview pipeline eliminates this. Rather than waiting for a full generation cycle to evaluate your result, you get a raytraced preview at interactive frame rates as you adjust parameters. The full generation only runs when you commit.
Infrastructure Specifications
| Specification | Value |
|---|---|
| GPU architecture | Custom cluster — NVIDIA H100 + Ada Lovelace hybrid |
| VRAM per render node | 80 GB HBM3 |
| Raytracing cores per node | 4th-generation RT cores |
| Cluster interconnect | NVLink 4.0, 900 GB/s bidirectional |
| Preview resolution | Up to 4K DCI (4096 × 2160) |
| Preview frame rate | 24fps sustained, 30fps burst |
| Latency (parameter adjustment → preview update) | < 80ms |
| Parallel user sessions | Autoscaling — no fixed limit |
The 80ms latency figure is measured end-to-end from parameter change to updated frame visible in the browser interface. This is below the human perception threshold for latency in video preview (approximately 100ms).
What Real-Time Raytracing Enables in Practice
Interactive lighting — adjust the direction, colour temperature, and intensity of your scene lighting and see the updated shadows, reflections, and subsurface response update in real time. No more generating multiple lighting variants and comparing static outputs.
Material iteration — change surface materials (matte to glossy, fabric to metal, skin tone variation) and see the specular and diffuse response update immediately.
Character positioning — move characters within the scene and see how the ambient occlusion, shadow casting, and inter-reflections update as you compose.
Environment matching — when compositing a generated character into a real-world background, the raytracer can match the environmental lighting in the background image and apply it to the generated subject in real time.
Access and Availability
Real-time raytracing preview is available to all Pro and Studio plan subscribers. It activates automatically in Vision Studio when your session is assigned a rendering node. No configuration is required.
Preview sessions are unlimited — there is no per-session billing. The raytracing infrastructure is part of the platform, not an add-on.
Generation (the final output step) continues to use the full generation pipeline for maximum quality. Raytracing is for preview and iteration only.