ANY VIDEO AT ANY RESOLUTION AND BANDWIDTH, IN THE BEST POSSIBLE QUALITY

VIDEO ENHANCEMENT AND GPU-BASED WORKFLOWS via SaaS

Get rid of noise and grain in your film and video content with our automatic and ultra-fast GPU-based low-delay denoising application. It performs strongly on RAW footage, and already compressed material can be processed effectively too. Get the right cleanup for your different distribution feeds and homogenize your VOD catalog, so your bandwidth ladder can be improved significantly. The picture quality of born-digital content and 35mm film will be optimized to allow ultra-low bitrates while preserving the highest visual quality.

Mastering the Image Chain: The IRIS-Denoising Ecosystem

At Wavelet Beam, we believe that achieving the highest visual quality means addressing noise where it matters most. Our IRIS-Denoising family is a precision toolset designed to integrate seamlessly into professional film and broadcast environments, from the first frame to the final stream.

1. Pre-Postproduction: The Clean Ingest

Before the creative work begins, the foundation must be perfect. We offer three specialized paths for early-stage optimization:

• IRIS.RAW (DNG2DNG): The gold standard. By denoising directly on Bayer-pattern data before de-bayering, we preserve the maximum dynamic range.
• BRAW2IRIS: Utilizing the BRAW SDK, this tool provides specific processing to optimize Blackmagic RAW footage for a clean start in the grading suite.
• IRIS (Universal): Our versatile solution for all standard video formats, ensuring every source is ready for the high demands of post-production.

2. Post-Postproduction: The Master Polish

Even after a clean ingest, intensive Postproduction, including heavy Color Grading or complex VFX, can stress the image signal. By applying IRIS to high-quality intermediate formats like FFV1 or ProRes, we eliminate newly introduced artifacts, ensuring a pristine Master File.

3. Delivery & Distribution: IRIS Per-Shot Encoding

The final stage is where efficiency meets excellence. Our IRIS Per-Shot Encoding technology revolutionizes the way content is delivered:

• Adaptive Intelligence: IRIS analyzes the master stream (e.g., from an FFV1 source) shot by shot.
• Optimized Compression: By understanding the unique noise profile and detail level of every individual shot, we enable codecs (H.264, HEVC, AV1) to work with surgical precision.
• The Result: Over 30% bitrate efficiency and maximum visual quality for VOD, Streaming, and Broadcast.

Modular by Design: Your Workflow, Your Choice

It is important to emphasize that IRIS does not need to be applied at every stage. Our ecosystem is built for maximum flexibility. Depending on your existing infrastructure or the specific technical nature of your signal, you can choose the most effective entry point:

• Source-Driven: Use IRIS.RAW or BRAW2IRIS to protect your RAW data before it enters the creative pipeline.
• Master-Driven: Use IRIS on an intermediate format like FFV1 to "clean up" after grading and VFX.
• Distribution-Driven: Implement IRIS Per-Shot Encoding to ensure the most efficient and high-quality delivery possible.

At Wavelet Beam, we provide the surgical tools for your image signal, but you decide where they are most effective for your specific production environment.

Per-Title, Per-Shot, and Content-Adaptive Encoding with IRIS.ANALYST

Not all video content is created equal, different scenes require different bitrates to deliver high-quality results. While we want to avoid wasting bandwidth, we also aim to preserve every detail. IRIS.ANALYST enables smarter encoding decisions by analyzing spatial and temporal complexity at the per-title and per-shot level.

Noise and grain are part of a video's spatial and temporal signal. As noted by Netflix, these factors alone can impact bitrate needs by up to 30%. At Wavelet Beam, we've confirmed this: leveraging IRIS.ANALYST can lead to a 30% reduction in bitrate, while maintaining or improving perceived visual quality.

IRIS.ANALYST Feature Set:

• Automatic selection of video noise profiles
• Scene and cut detection for per-shot encoding
• Measurement of video resolution, no test target required
• Quantified video noise levels using PSNR [dB]
• Spatial and temporal complexity measurement
• Contrast and brightness evaluation

Turn on AUDIO!

The LA Downtown and Tiger clips above are zoomed and scaled from 9K to UHD (shot on 9x7).

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!

Wavelet Beam proudly presents IRIS.AI: Advanced AI-Based Video Upscaling

Before applying AI-driven upscaling, it's critical to first denoise the video. Why? Because noise in the original footage can confuse AI algorithms, leading to artifacts and reduced visual quality in the final result.

IRIS.AI combines Wavelet Beam's cutting-edge video denoising with powerful deep learning upscaling to ensure exceptional results. Clean input means the AI can focus on real image content, not on random noise, resulting in sharper, more natural details and improved resolution.

By preprocessing with IRIS, IRIS.AI receives a cleaner signal, allowing the upscaler to add pixels more accurately. The result: enhanced visual quality, higher perceived resolution, and a stunning viewing experience, ideal for remastering, broadcasting, or archive restoration.

Revolutionizing the BRAW Workflow: From Budget Cameras to 12K or 16K Immersive

At Wavelet Beam, we believe that image quality should not be limited by your hardware's price tag. Today, we are proud to announce the B2B Early Access phase for BRAW2IRIS, our fully automated, high-fidelity noise management solution.

"Start grading again instead of waiting for the denoising."

Why settle for "good" when you can achieve professional-grade results from any Blackmagic sensor? BRAW2IRIS transforms your footage into a cinematic masterpiece by integrating our advanced IRIS-Denoising directly into your existing BRAW pipeline.

The Technical Advantage:

• Near-Real-Time Speed: Speed is our standard. Our process is so optimized that your RAID speed is the only limitation, not the processing.
• 16-Bit Precision: We process your 12-bit source in a dedicated 16-bit pipeline to eliminate rounding errors and banding.
• Maximized Dynamic Range: By cleaning the noise floor, we unlock deeper shadows and superior HDR results, giving you more stops of usable dynamic range.
• Higher Usable Resolution: We reveal fine details previously masked by digital noise, effectively increasing your final output clarity.
• Organic Cinematic Texture: We don't "smear" pixels. We preserve a natural noise floor while removing distracting artifacts, maintaining that sought-after filmic look.

Scalable from Indie to 12K or 16K Immersive

Whether you are managing a fleet of budget-friendly Blackmagic cameras or pushing the boundaries of 12K or 16K Immersive productions, BRAW2IRIS scales to your needs, delivering superior image stability and professional-grade master files, fully automated.

B2B Early Access

We are currently rolling out BRAW2IRIS exclusively to select production houses and B2B partners who are ready to push their BRAW workflow to the absolute limit.

Are you ready to elevate your footage and reclaim your time? Contact us at: info@waveletbeam.com

#WaveletBeam #BRAW2IRIS #BlackmagicDesign #PostProduction #16KImmersive #HDR #Cinematography #HighFidelity #BRAW #IRIS #QVBE #VMAF #PSNR #12K

Wavelet Beam proudly presents IRIS.RAW: The Next-Generation Video RAW Data Denoising Solution

Wavelet Beam introduces IRIS.RAW, a powerful solution for video RAW data denoising. Modern digital cameras use single-chip Bayer sensors that capture only one color channel per pixel. In postproduction, the two missing color channels must be interpolated through the de-bayering process. However, noise inherent in the camera signal can distort this process, resulting in inaccuracies.

By reducing noise levels before de-bayering, IRIS.RAW ensures much higher resolution and improved HDR performance, delivering pristine quality for your video footage. Whether you are working in 4K, 6K, or even 12K, IRIS.RAW provides a clean foundation that enhances both the editing and grading processes.

Our solutions, IRIS and IRIS.RAW, are GPU-accelerated, high-fidelity denoising applications that offer one of the fastest processing speeds on the market. They are designed to integrate seamlessly into high-resolution workflows where conventional tools struggle to handle real-time denoising.

IRIS.RAW is the perfect solution for workflows requiring real-time denoising, especially when handling high-resolution content that demands significant CPU/GPU power. By offloading the denoising process, you free up resources for color grading and real-time video enhancement, ensuring better dynamic range, resolution, and overall streaming performance.

Based on Wavelet Beam's GRAIN & NOISE REDUCER, IRIS.RAW extends our commitment to high-quality video processing and seamless workflow integration.

Wavelet Beam Services offers an all-inclusive package for cleaning video content and film restoration. Nowadays, the number of UHD television sets is increasing. Thus, the quality of DVD and Blu-ray products also has to be increased to provide the best viewing experience for customers. Wavelet Beam Services offers noise management that takes your film material to the next level. Be ready for UHD.

	PROFESSIONAL SERVICES: Hands-on training and consultancy services for FFmpeg FFmpeg is a free open-source software project consisting of a suite of libraries and programs for handling audio and video. FFmpeg serves as the engine for most of the largest cloud encoding farms in the world and gives you an alternative to commercial transcoding products. Possible topics: 1. Quality matrix: VMAF and PSNR 2. Optimizing your bitrate ladder 3. FFmpeg automation on LINUX 4. Speedup encoding for Adaptive Bitrate Encoding 5. How to analyze files with MediaInfo, Bitrate Viewer, Apple's AVQT 6. Using FFv1 as Intermediate Format
	PROFESSIONAL SERVICES: Video Encoder Comparison using Video Quality Metrics As an independent consultant, Wavelet Beam offers services for video encoder shootouts and optimization of video bitrate ladders. These tasks are time-consuming and the measurement of the video quality metrics needs a lot of compute power, which Wavelet Beam is also offering. In the design phase of a new video workflow, it is important to know all details about encoding speed and video quality in advance. Additionally, we offer video test sequences, which make it possible to measure parameters like resolution or dynamic range of the End-to-End workflow.
	Perceptually Optimized Video Coding and Quality Measurement In recent years, Wavelet Beam has developed a new, high-fidelity noise management system. All our image analysis and enhancement technologies are based on the underlying model of noise and signal shares. This knowledge is what we are using in our Perceptual Video Quality Measurement process. As we learned, a lot of codec vendors don't like to compare their encoders because real numbers would be a risk to sales. If you have a vendor shootout, Wavelet Beam Services will provide you with reliable numbers. If you are a codec vendor, the Wavelet Beam Perceptually Optimized Video Coding SDK is a great opportunity to be two steps ahead when it comes to picture quality and ultra-low bandwidth encoding.
	PROFESSIONAL SERVICES: TEST CHARTS Collaboration between Cinelab London and Wavelet Beam Vendor-independent 35mm film test strips are available again INCLUDING: * 35mm test charts * Test chart generation * Image quality analyses * Workflow analyses The technical quality parameters of film scanners are changing over time. The sensor performance is decreasing and artifacts such as dead or hot pixels will appear. Noise levels can also increase over time and the later you recognize quality issues, the more time and money you lose. In a collaboration between Cinelab London and Wavelet Beam, we offer 35mm negative and print film. Additionally, Wavelet Beam offers the analysis of the scanned film material.
	PROFESSIONAL SERVICES FOR GPU WORKFLOWS INCLUDING: * Project Management * Implementation * CUDA * Image Processing LIBs * Algorithm Development * Deep Learning for Video Analytics
	PROFESSIONAL SERVICES FOR DTV AND FILM INCLUDING: * Project management * Implementation * System analyses * RFQs * Test chart generation * Image quality analyses

IRIS FOR AI: BROADCAST, EDGE-AI & MACHINE VISION

Clean signals for AI models in live broadcast, AI cameras, CCTV analytics and autonomous systems

New Article, September 26, 2026

Why do AI video models in broadcast environments often fail when moving from testing to live transmission?

Broadcasters operate with an unpredictable mix of camera sources: studio setups, field cameras, drone feeds, archives and user-generated content. Every source brings its own sensor noise, gain artifacts and heavy video compression.

During AI model training, neural networks fall into shortcut learning. Instead of analyzing real scene content, they learn specific sensor noise signatures or compression artifacts. As a result, when an automated camera tracking or live object detection system encounters a different camera feed, its detection confidence drops immediately.

Standard denoising solutions fail in broadcast because they simply blur or smooth the picture. Blurring wipes out fine textures, sharp edges and essential detail, stripping away the exact information AI models need to make accurate predictions. At Wavelet Beam, we solve this bottleneck for broadcast pipelines using IRIS-Denoising.

Why IRIS-Denoising Changes the Game for Broadcast AI

• Signal extraction without blurring: IRIS-Denoising does not smear or flatten video data. It isolates and lifts the true visual signal straight out of the noise. Fine details, high-frequency textures and crisp edges are fully preserved, delivering clean, high-contrast feature maps for neural networks.
• Fully automatic across any broadcast feed: Broadcast engineers do not have time to manually tune noise profiles during live productions. IRIS Video-Denoising adapts dynamically to changing ISO levels, low-light gain or aggressive codec compression across every incoming stream.
• Higher inference accuracy for automated workflows: Whether for live player tracking, automated graphic overlays or AI-driven stream moderation, cleaner input signals directly improve bounding box precision, tracking continuity and overall confidence scores.
• Built for live broadcast infrastructures: IRIS-Denoising operates seamlessly in both file-based offline environments and live workflows via MXL and DMF integration.

At Wavelet Beam, we are also building AI-based Media Functions that integrate directly into live broadcast pipelines, enabling high-precision AI processing with minimal latency.

Read and discuss this article on LinkedIn

#IRISDenoising #WaveletBeam #BroadcastEngineering #LiveProduction #ComputerVision #VideoAI #BroadcastAI #DMF #MXL #IRIS

Edge-AI, Machine Vision and Video Coding for Machines

The paradigm shift from human-centric to machine-centric video encoding starts with one unavoidable prerequisite: a clean signal. That is exactly what IRIS delivers.

The Shift: From Human Perception to Machine Inference

For decades, video encoding was optimized for human viewers: PSNR for mathematical fidelity, SSIM and VMAF for perceptual quality. Artificial intelligence has fundamentally different requirements. It does not care whether a picture looks beautiful. It needs feature preservation, temporal consistency, inference-relevant frequencies and maximum information density.

Where IRIS Sits in the VCM Stack

Video Coding for Machines (VCM) defines a four-layer architecture. Many VCM solutions assume clean input and start at layer 2 or 3. IRIS owns Layer 1, Signal Conditioning, the layer that sets the ceiling for everything above it. IRIS.ANALYST extends this into Layer 2 with per-shot complexity metadata that machine-optimized encoders can act on directly.

The Edge-AI Pipeline Problem

A modern CCTV deployment produces 4K video 24/7 across hundreds of cameras. 99.9% of that footage contains no relevant event. The industry is shifting from central-server analysis of raw streams toward edge-based inference, transmitting only events, metadata and compressed reference video. IRIS improves inference quality at the edge while simultaneously reducing the bitrate of any reference stream that must be transmitted.

Relevant Market Segments

CCTV and video analytics platforms · Edge-AI SoC and camera silicon vendors · Video Management Systems · Cloud video analytics · Streaming and CDN encoding infrastructure · Autonomous systems (robotics, ADAS, UAV) · MPEG VCM standardization working groups

IRIS & MXL

IRIS Video Denoising as a native MXL Media Function

Press Release

Dynamic Media Facility & MXL: A Fundamental Architectural Step

Paderborn, August 30, 2026: The demands placed on modern media production keep growing: dynamic formats, variable data rates and the pressure to achieve high cost efficiency all call for flexible infrastructures. While classic setups depend on constantly exchanging video data between individual devices over physical or IP interfaces such as SDI and ST 2110, even for internal processing steps, the Dynamic Media Facility (DMF) in combination with the Media eXchange Layer (MXL) introduces a fundamental architectural step.

As a Software-Defined Video (SDV) solution built on Kubernetes, it decouples video processing from dedicated hardware. Established standards such as SDI and ST 2110 fully retain their role for external signal connectivity, while the internal data exchange between the processing functions takes place highly efficiently and directly within the software infrastructure.

Technical Context

DMF is the EBU/NABA reference architecture: NMOS decides, MXL transports the data, Kubernetes schedules the compute.

From Static Hardware Standby to Dynamic Software Resilience

In traditional broadcast environments, safeguarding operations relies on substantial over-provisioning: dedicated hardware cards (FPGAs) and standby devices have to be kept available at all times to guarantee 1:1 or 1:N redundancy, capacity that ties up capital while sitting idle in regular operation and that is tied to the replacement cycles of the hardware.

The DMF architecture resolves this problem at its root:

• Hardware decoupling with a hybrid option: Instead of device-bound hardware, containerised software on standard servers (COTS) primarily handles signal processing. This does not rule out specialised compute at all: GPUs as well as existing FPGAs remain important building blocks of modern workflows.
• A call to broadcasters and technology providers: To ensure optimal integration, dedicated GPU and FPGA interfaces now need to be actively contributed to the MXL standard. Broadcasters and technology providers are explicitly invited to formulate their concrete use cases and requirements, so that these interfaces can be shaped for real-world practice.
• Elastic redundancy instead of idle hardware: Redundancy is orchestrated flexibly through Kubernetes. If an instance fails, the system automatically scales new workloads in real time. Permanently maintaining dedicated standby devices is no longer necessary.
• High interoperability & in-memory performance: Through the MXL SDK, software applications from different vendors exchange uncompressed media data directly via shared memory with extremely low latency. Broadcast customers thereby save a large share of the costly integration effort.

The Best of Both Worlds: An Efficient Hybrid Cloud Model

A key characteristic of the overall architecture is its intelligent hybrid cloud concept:

• Base load covered on premises: The 365-day base load stays cost-efficiently in your own local data center.
• Dynamic cloud scaling: Live events with extreme load peaks, such as sports coverage, are offloaded flexibly to the cloud.
• Bandwidth optimization & image quality: Thanks to the reduced bandwidth requirements of the intermediate format, the system combines top technical performance during the transfer to the cloud with significant savings. This guarantees the best possible conditions for outstanding distribution image quality at minimal infrastructure cost.

Cloud-Native Advantages at a Glance

• Maximum scalability: Container orchestration allows processing power to be adjusted dynamically to the production workload.
• Simplified maintenance: Continuous deployment and automated updates reduce downtime and administrative effort to a minimum.
• Sustainable investment protection: Unlike device-bound hardware, the software logic is not tied to fixed replacement cycles; continuous updates keep it permanently up to date.

Winner of the IBC2026 Innovation Awards

As an active contributor from Wavelet Beam to this forward-looking technology approach, we are delighted that the open-source Media eXchange Layer (MXL) project has won the renowned IBC2026 Innovation Award in the Content Creation category. A huge thank you to the EBU, CBC/Radio-Canada, NABA, the Linux Foundation, NVIDIA and all participating partners. This recognition underlines the pioneering significance of MXL and DMF for the entire broadcast industry.

Current Implementation

The current version operates with Host Memory, delivering full IRIS denoising quality within the MXL pipeline.

Next step: Direct integration with the MXL Device Memory Ringbuffer, removing the host memory round-trip and unlocking the full performance potential for GPU-native, zero-copy denoising at scale.

About Wavelet Beam

Wavelet Beam is a leading specialist in GPU-accelerated video processing technologies. As an active contributor, Wavelet Beam helps shape forward-looking open-source initiatives and standards for software-defined media infrastructures.

A Leap in Efficiency with IRIS: Major Savings in Bandwidth and Bitrate

As a central building block in modern software-defined video pipelines, Wavelet Beam’s core technology IRIS comes into play. IRIS resolves one of the biggest bottlenecks in video distribution: disruptive image noise and grain:

• Precise noise management: IRIS separates unwanted noise from genuine image detail with high precision and lifts the signal out of the noise floor.
• Massive bitrate savings: By reducing erroneous motion vectors and visual disturbances, encoders have to process considerably less redundant data. This enables bitrate savings of up to 30 % and more, at the same or even improved perceived visual quality.
• Optimal distribution quality: Whether live broadcast, streaming or archive VOD: IRIS delivers clean master signals and creates the perfect basis for maximum compression efficiency along the delivery path.

Wavelet Beam has implemented a first version of IRIS Video Denoising using the MXL SDK. IRIS runs as a standalone MXL Media Function inside a Docker container, fully integrated into the MXL infrastructure and ready to process live video streams with GPU-accelerated noise management.

MXL Workflow

Infrastructure

The deployment comprises five MXL applications, each running as a standalone Media Function in its own Docker container:

• SDI to MXL Bridge: Ingests SDI video streams via the Blackmagic DeckLink SDK and exposes them natively as MXL flows.
• IRIS-Denoising: GPU-accelerated real-time processing. The container encapsulates the complete IRIS denoising pipeline and cleans up live sources at the highest image quality, before intermediate formats for post-production are generated.
• GPU UHD Video Recorder: GPU-accelerated recording of MXL flows up to UHD resolution.
• 10-bit Uncompressed Viewer Application: For presenting uncompressed media pipelines in real time.
• Video Player: Real-time playback of MXL flows inside the fabric.

All applications share a common licensing component:

• CodeMeter License Server (VM): A dedicated virtual machine hosts the Wibu CodeMeter 8.40 service, providing licenses over the network via CmNet (Port 22350). Each container authenticates against this server on every start.

ARTICLES & POSTS

Every LinkedIn article and post from Dirk Hildebrandt, in one searchable archive