Nokia has open-sourced its Neural Network Post Filter (NNPF), bringing the first AI-based video enhancement technology into international video standards and enabling higher-quality streaming with existing codecs instead of proprietary solutions.
Nokia has open-sourced its Neural Network Post Filter (NNPF), an AI-powered implementation that enhances compressed video quality after decoding on compatible consumer devices. According to Nokia, it is the first AI-based video enhancement technology to be integrated into international video standards.
NNPF applies a neural-network post-processing filter to recover visual details lost during video compression. Rather than introducing a new codec, it works alongside existing standards such as High Efficiency Video Coding (HEVC) and Versatile Video Coding (VVC) using Versatile Supplemental Enhancement Information (VSEI) and Supplemental Enhancement Information (SEI) metadata to identify and apply the appropriate neural-network model during playback.
Nokia researchers Miska Hannuksela and Jill Boyce contributed to the Versatile Supplemental Enhancement Information standard, which defines how this metadata is carried with coded video.
If widely adopted, the technology could help streaming providers improve perceived video quality, reduce video CDN costs and make better use of AI acceleration already available in modern consumer devices, without abandoning established compression formats.
Deployment will still require support from encoding platform providers, chipmakers, device manufacturers and video application developers.
The release also highlights the contrast between open and proprietary AI video enhancement approaches. While NVIDIA’s AI enhancement is tied to RTX GPUs and restrictive SDK licensing, Intel’s OpenVINO demonstrates growing support for open AI inference across CPUs, GPUs and NPUs.
