WaveletBeam IRIS Analyst

IRIS.Broadcast: Real-Time Denoising for Broadcast-Grade Video

IRIS.Broadcast is a high-performance, GPU-powered Digital Video Noise Reducer (DVNR) designed for live production and broadcast environments. Scale from 8+ simultaneous 1080p streams on a single GPU to multi-GPU UHD and 8K pipelines, all with ultra-low delay.

IRIS not only delivers visually clean output, but also reduces video bitrate by up to 30% by optimizing motion vector precision.

  • GPU-accelerated, highly scalable
  • Real-time 1080p, UHD, and up to 8K
  • Multi-stream efficiency, ideal for high-density workflows

At Wavelet Beam, we're pushing the boundaries of what's possible in video technology. We're excited to introduce real-time support for IRIS and IRIS.Analyst.

Our advanced IRIS SDK powers Video Noise Management and Video Analytics at scale, enabling broadcasters to reduce bitrates while enhancing video quality, delivering a richer, sharper viewing experience.

What sets IRIS apart? It's not just about reducing visible noise, it's about lifting the signal out of the noise floor. This means potentially recovering image information that would otherwise be lost or unusable, far surpassing traditional noise reduction methods that often blur or distort fine details.

Powered by CUDA acceleration, IRIS is built for performance and scale. It can process multiple video streams on a single GPU, and scale efficiently across multiple GPUs, making it ideal for high-demand broadcast environments.

A demo setup is available for interested B2B partners. Currently supporting SDI, with ST 2110 integration coming soon.

Let's redefine the future of broadcast, together.

This video has sound!

IRIS SDK

Results and Insights

Measurements and findings from our work, summarized from Dirk Hildebrandt's posts and articles.

  1. Up to 8K at 60 fps on NVIDIA GPUs

    Through massive CUDA parallelization, IRIS.Broadcast processes the most demanding signals up to 8K at 60 fps directly on NVIDIA hardware, with high-precision noise reduction and no loss of time.

    Read on LinkedIn

  2. Multiple UHD streams per GPU

    Using NVIDIA GPUs and CUDA, IRIS.Broadcast processes multiple UHD streams at 50 fps simultaneously on a single GPU in real time, for everything from live sports to cinematic productions.

    Read on LinkedIn

  3. Why GPU-accelerated DVNR matters for broadcasters

    Noise compromises visual quality and hinders compression efficiency, which drives up distribution costs. The GPU-based DVNR, powered by CUDA and a scalable HPC architecture, removes noise without sacrificing detail, for live feeds and archived footage alike.

    Read on LinkedIn

  4. Real-time UHD on a single GPU since 2016

    With the NVIDIA Pascal generation, high-fidelity grain and noise management became available for real-time UHD processing on a single GPU, including HDR footage.

    Read on LinkedIn

  5. Noise, motion vectors and statistical multiplexing

    Noise reduces temporal redundancy and can be mistaken for motion, which leads to faulty motion vectors. In a statistical multiplex, a noisy service needs more bandwidth and takes it away from the other services, so one noisy channel can lower the quality of several.

    Read on LinkedIn

Frequently Asked Questions

What is IRIS.Broadcast?

A GPU-powered Digital Video Noise Reducer (DVNR) for live production that cleans signals from camera sensors or remote feeds directly in the signal path, before they reach the encoder.

How many streams can one GPU handle?

8+ simultaneous 1080p streams on a single GPU, scaling to multi-GPU UHD and 8K pipelines, all with ultra-low delay.

Does live denoising reduce bitrate?

Yes. By improving motion vector precision, IRIS.Broadcast reduces the video bitrate by up to 30%.

Up to which resolution and frame rate does IRIS.Broadcast work in real time?

IRIS.Broadcast processes signals up to 8K at 60 fps on NVIDIA GPUs. A single GPU handles 8+ simultaneous 1080p streams, larger setups scale to multi-GPU UHD and 8K pipelines.

Why does denoising matter in a statistical multiplex?

A noisy service needs more bandwidth and takes it away from the other services in the multiplex. Removing noise before encoding keeps the bitrate demand of each service lower and protects the quality of the whole multiplex.