• How our latest progress in developing and standardizing neural network-based post filtering (NNPF) enables real-time video quality enhancements across multiple codecs without increasing the bitrate. 
• This innovation will enable viewers to experience high-quality video streaming anywhere, without needing new or expensive gear. 

For video to truly immerse its audience, it must preserve the details that make images lifelike. Every frame needs to carry the depth, color, and clarity the creator intended. 

Yet streaming often strips those details away. To keep files small enough to travel quickly across networks and play smoothly on any device, some compression algorithms may remove visual information that feels less essential. The result is banding in skies, textures that look flat, and shadows that lose their subtlety – the little things that make images feel real. 

Today, Nokia reveals its progress in developing neural network-based post filtering (NNPF), restoring those lost details by enabling real-time video quality enhancements across multiple codecs. In short, NNPF brings AI-powered perceptual optimization to video delivery, enhancing quality without increasing bitrate – and doing so in a way that is modular, scalable, and future-proof.

From compression to clarity 

Imagine watching a new release at home and seeing every detail, from the glint of light in an actor’s eye to the texture of a distant landscape, with the same clarity and depth you’d expect in a cinema.

This leap is enabled by Nokia’s neural network-based post filtering. Unlike traditional approaches, NNPFs can be applied across codecs and updated on a per-video sequence basis using the developed Versatile Supplemental Enhancement Information (VSEI) standard. That means video quality can be improved instantly, without re-encoding existing material. The result is richer colours, sharper textures and smoother gradients, all delivered on the devices people already own. This represents the first time AI has been formally integrated into a video coding standard like VVC/H.266. 

Neural network hardware support is now making this possible at scale. Dedicated processors for NN execution – from Tensor Processing Units to Neural Processing Units – are already built into laptops, mobile phones, and even TVs. By targeting this hardware, NN-enabled tools for video quality improvements can be applied widely, reaching millions of consumer devices.

Driving this progress is a Nokia research team focused on neural video processing, immersive media, codec optimisation and real-time rendering – creating the innovations that will shape the next generation of ultra-high-definition multimedia experiences.

Experience the innovation: IBC2025

Later this month, Nokia will provide visitors at IBC2025 with a first-hand experience showcasing the real time inference of NNPF for various codecs, including VVC real time decoding, on everyday consumer devices. 

Attendees participating in the demonstration, available in Hall 5 – IBC Context Everywhere – 5.H60 will see the effect of neural network post filters to increase video quality, especially regarding banding artefacts removal. 

Nokia will also be demonstrating volumetric video communication capabilities, showing a free viewpoint six-degrees-of-freedom volumetric video on Xreal 2 Air Ultra AR Glasses. The demo will showcase volumetric video’s abilities to immerse users into experiences where they can consume real-world live captured content as if they were physically there. 

Nokia’s leadership in multimedia innovation

Over the past three decades, Nokia has made significant contributions to advancing multimedia technologies, creating more than 5,000 multimedia inventions. 

We have also been among the most active contributors to the VSEI standardization –holding editor positions in the latest version of the VSEI standard. We pioneered content-adaptive fine-tuning techniques for neural networks used in quality-enhancement post-filters, with the aim of enabling video services and devices to improve subjective video quality through the VSEI NNPF, supported by a strong ecosystem.

The progress revealed today with neural network post filtering is one element of Nokia’s wider leadership in multimedia innovation. By combining expertise in AI, video compression, and immersive media, we’re working to shape the next generation of digital experiences – unlocking the full potential of video, ensuring it remains as vivid, efficient, and immersive as the world it reflects.