V-Nova VC-6 AI Blueprint enables multi-model inference from single streams

This development addresses the high computational overhead of real-time video analytics by moving intelligence into the codec layer. By decoding only the specific pixels required for inference, streaming providers can significantly reduce memory bandwidth and I/O requirements on resource-constrained edge devices. This shift from full-frame processing to adaptive, region-based decoding aligns with the industry's move toward more sustainable and scalable AI-driven monitoring. As multi-model pipelines become standard for surveillance and forensic replay, the ability to maintain high throughput without sacrificing detection accuracy will be a critical differentiator. Watch for whether other codec standards adopt similar hierarchical metadata structures to compete with VC-6 in automated vision workflows.

V-Nova's VC-6 codec has been building institutional credibility since its formal standardization. SMPTE published VC-6 as the ST-2117 standard following a comprehensive peer review process, confirming the codec's AI-driven hierarchical compression approach that avoids traditional DCT or wavelet transforms. The codec's architecture, based on S-Trees, enables multi-resolution and region-of-interest decoding natively, capabilities that the VC-6 AI Blueprint now extends into multi-model inference workflows. V-Nova has positioned VC-6 as a mezzanine codec optimized for contribution, editing, and archival applications, with early deployments through its P.Link product for remote production.

The broader V-Nova ecosystem continues to expand across broadcast standards. V-Nova's LCEVC technology was adopted as part of Brazil's DTV+ national standard, which layers LCEVC as a residual correction on top of VVC within an ATSC 3.0 transport framework. The ATSC itself formalized VVC and LCEVC support when it approved the A/345 VVC Video standard on July 17, 2025, including optional scalability based on LCEVC aligned with CMAF. V-Nova and Ateme demonstrated live LCEVC-enhanced 4K broadcast over ATSC 3.0 at the NextGen Broadcast Conference in 2025, proving interoperability of the full encoding and transmission chain. Fabio Murra, V-Nova's Senior Vice President of Product and Marketing, played a key role in both the TV 3.0 ratification and the advancement of SMPTE VC-6 across software and hardware implementations.

On the technical side, VC-6's hierarchical structure delivers measurable efficiency gains that underpin the AI Blueprint's value proposition. V-Nova reported that VC-6 achieves up to 60% reduction in disc access for 4K files and 2-4x faster decoding than existing ProRes-based solutions, while 8K processing requires only 12.5% of full resolution to be demosaiced. These figures directly support the AI Blueprint's claim that Region-of-Interest decoding can reduce memory bandwidth for edge inference pipelines. The codec's compatibility with major container formats including MXF and QuickTime means it integrates into existing production and analytics workflows without re-encoding overhead, a practical advantage for specialized silicon architectures running models like YOLOv5 or MediaPipe FaceMesh.