SMPTE has now fully ratified and published V-Nova’s new video codec, VC-6, following a peer review process. Published as SMPTE ST-2117, VC-6 is one of the first video codecs to use AI as part of its approach to compression. Instead of applying transforms such as DCT or wavelet, VC-6 employs a hierarchical series of S-Tree data structures, and achieves its compression by reusing the same structures at different resolutions. This capability makes the codec very fast when implemented in software.
The AI-powered hierarchical approach means that full-resolution proxies are intrinsically available – that is, as part of the format – as is region-of-interest decoding. This functionality reduces demand for data transfer and decode processing. Also, VC-6’s compatibility with the major industry container formats like MXF and QuickTime means it can be directly incorporated into most workflows.
The codec has the efficiency needed to work with 4K and 8K media in production. In particular, the emerging 8K format has associated production challenges that VC-6 can help overcome. For example, users can capture video in 8K from RAW without having to cross-convert to other resolutions at any time later in the process. Instead, VC-6 can output proxies automatically and enable editing at multiple resolutions all from the single file.
V-Nova says that initial use cases for their new codec show marked reductions in 8K processing by requiring only 12.5% of the full resolution to be demosaiced, up to 60% reduction of disc access for 4K files, and decoding that is 2 to 4 times faster than existing ProRes-based systems.
VC-6 is shipping now as part of V-Nova’s P.Link dual UHD/8HD video encoder/decoder device for contribution and remote production applications. The flexibility of the software supports high density HD and UHD bi-directional processing. Its Dynamic Multiplex allows the encoder to adjust the bandwidth between services according to complexity and priority, in real time, allowing more services or higher quality for a given link.
The company is currently working on a multi-platform SDK and cloud service deployment of VC-6, expected to be made available via an Early Access Program later this year (2020). www.v-nova.com