How do you render volumetric clouds in real time for broadcast TV and keep the visual quality of pre-rendered content? How can you run an AI avatar without spinning up a new virtual machine for every user? These are the kinds of technical problems that drive Pekka Varis and his team at Cineshare. From their studio in Finland, they’re developing new approaches to character animation and virtual production that balance visual quality with performance constraints.
Solving Real-Time Weather Visualization
Cineshare faced a specific challenge on behalf of Finland’s national broadcaster, YLE: creating live weather visualizations during broadcasts. The technical pipeline starts in Embergen, where the team generates cloud formations and exports them as VDB (Volume Database) files.
“VDB files give us efficient storage for sparse volumetric data, but they’re complex to work with,” explains Varis, “Each cloud formation can be several gigabytes. You need substantial GPU memory just to load them, let alone manipulate them in real-time.”
Cineshare’s breakthrough came from NVIDIA Omniverse’s Universal Scene Description (USD) framework, which lets multiple artists collaborate on scenes simultaneously. Using their Dell Precision 3680 workstation with an NVIDIA RTX 6000 Ada GPU, the team could layer changes from different tools and artists in real-time. While one artist refined volumetric lighting, another could work on particle systems for rain and atmospheric effects.
“NVIDIA Omniverse transformed how we handle these complex simulations Before, we’d be passing huge files back and forth between different tools. Now, everything updates live. We can actually refine lighting during client reviews instead of waiting for overnight renders.” – Pekka Varis, Cineshare Co-Founder & Creative Director
Adapting to VR
NVIDIA Omniverse running on Dell Precision workstations, streamlined collaboration on 3D simulations and renders, but the team faced a new challenge when YLE, Finland’s national broadcaster, requested a VR adaptation for public demonstrations. VR demands consistent high frame rates to prevent motion sickness says Varis. We spent three months optimizing our real-time pipeline, and now visitors can fly through storm systems in VR while experiencing high visual quality with dynamic wind and rain. The seamless integration between Embergen, Unreal Engine, Varjo VR support made this possible.
Character Creation and Real-Time Digital Humans
Weather visualization pushed real-time graphics’ boundaries for Cineshare. However, the studio’s primary focus remains on character animation. As one of the first Omniverse ambassadors, Varis pioneered workflows between character creation tools and real-time rendering.
“I got involved with Omniverse early because I saw its potential for character work,” Varis explains. “The USD framework lets us maintain complex character rigs and animations across different tools. We can build a character in one application, animate it in another, and see everything update live in Omniverse.”
With Precision workstations equipped with NVIDIA professional graphic cards, the studio now achieves 120 FPS with their SSS (subsurface scattering) skin shader implementation running through Omniverse. “Rendering characters in real time affects more than speed though” says Varis. “Now, we can manage dynamic hair and lifelike skin details that used to require offline rendering. Omniverse lets us iterate on these effects instantly.”
Pushing AI Integration Further
Cineshare’s latest technical challenge involves integrating NVIDIA’s Avatar Cloud Engine (ACE) with their character pipeline. The University of Helsinki recently deployed an ACE avatar through Unreal Engine’s Pixel Streaming, revealing both opportunities and challenges.
“Current ACE deployments need a dedicated GPU instance per user,” Varis explains. “Every time someone wants to interact with the avatar, a new virtual machine spins up in the cloud. It works, but it’s expensive and complicated to maintain.”
This led Cineshare to explore running large language models directly on-prem with workstation GPUs. Using the RTX 6000 Ada’s 48GB of VRAM, they can run Llama models locally on their Dell Precision workstations for avatar interactions and technical assistance. The team is now working to integrate these local AI capabilities with their Omniverse character pipeline.
“Local AI processing gives us more control,” notes Varis. “We’re building tools to help artists navigate complex technical systems – things like Linux environments and command-line interfaces that traditionally create bottlenecks in production pipelines. The AI provides real-time guidance without cloud dependencies or latency issues, and it boosts up the AI Enterprise customer support I trust on building these experiences.”
Looking Ahead
For Cineshare, the future lies in combining these technologies: real-time rendering, collaborative workflows through Omniverse, and local AI processing on Dell hardware. “Each project pushes us to solve new technical problems,” says Varis. “But having a unified platform in Omniverse lets us build on our solutions instead of starting from scratch every time.”
The Cineshare team continues to develop new approaches to character animation and virtual production, balancing visual quality with performance requirements. Their work demonstrates how modern GPU hardware and collaborative platforms like Omniverse are transforming what’s possible in real-time graphics.
They have also just launched their first SaaS product, Omniverse Easy Start!
Engineered for unmatched performance, Dell Pro Max high-performance PCs* harness the full potential of professional-grade graphics. To learn more about the benefits of high-performance PCs and NVIDIA RTX GPUs, click here.
*Dell Pro Max high-performance PCs, previously referred to as Dell Precision workstations
