V-Nova is advocating for hierarchical, compute-aware data formats to significantly improve the efficiency of Vision AI processing. These formats, designed to be AI-native, aim to reduce compute power, energy consumption, and latency by allowing AI to access visual data with purpose, rather than processing everything.
The streaming industry is pivoting from video for human consumption to video for machine analysis, yet legacy codecs like H.264 remain a fundamental bottleneck for AI efficiency. V-Nova's push for compute-aware formats addresses the critical 'data-to-tensor gap,' where expensive hardware sits underutilized due to unstructured data pipelines. For operators, this offers a path to scale computer vision and real-time metadata generation without a linear increase in cloud compute costs or energy footprints. Watch for the integration of these hierarchical structures into the upcoming ISO/IEC 23888 (MPEG-AI) standards for Video Coding for Machines.
V-Nova has been building a public case for hierarchical, compute-aware data formats across multiple industry venues in 2026. At the Mile High Video conference in Denver, V-Nova and Intel's Edge Computing Group presented joint benchmark results showing that in a conventional 4K analytics pipeline using SSD MobileNet V2, only approximately 1.4% of decoded pixels were ultimately used for inference, with the remainder consuming decode cycles, scaling effort, and memory bandwidth. The same presentation reported that SMPTE VC-6's selective resolution and frame access reduced processing time by up to 96% on the VisDrone UHD dataset, while MPEG-5 LCEVC base-layer decoding cut CPU execution time by 33 to 52% and improved GPU throughput by roughly 3x.
Ahead of NAB Show 2026, V-Nova SVP of Strategic Analytics Guendalina Cobianchi published a pre-show argument that today's video formats were designed for human eyes, not for machines, framing three conversations for the codec industry: compute-aware data formats, GPU-accelerated visual AI in live production and archive workflows, and the swXtch AI Router for distributing AI inference at broadcast scale. The argument carries practical consequences for equipment buyers, since encoding and transport formats optimized for perceptual quality may not serve AI workloads that consume media as structured data rather than continuous playback.
V-Nova has also detailed how SMPTE VC-6's architecture differs from legacy alternatives at a technical level. The format encodes every frame as a multi-resolution pyramid with up to 255 independent data planes, supporting region-of-interest decoding and selective bitstream retrieval without requiring specialized silicon. The company has paired VC-6 with swXtch.io's multicast network layer to demonstrate live production workflows where a single VC-6 stream serves dozens of parallel AI workloads, each subscribing to the specific level of quality and region of interest it needs, eliminating the need for proxy encodes or duplicated feeds.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source