D2C studios, broadcasters and platform operators can expect a 20% encoding efficiency boost thanks to the use of AI – in some cases helping them delay codec upgrades beyond HEVC thanks to short-term reductions in CDN costs. Video processing specialist Media Excel is promising the gains with its new DIVA (Dynamic Intelligent Video Adaptive) encoding technology, which will be launched at IBC in September and become generally available soon after.
The new DIVA solution, which leverages deep learning models, is focused on streamed HEVC-encoded content. Media companies benefit from the usual use-case possibilities associated with compression improvements: better video quality within the same bandwidth or the same video quality using lower bit rates, with the latter also allowing reach into the deeper corners of wired and wireless networks.
Right now, it is cost reductions that are the most pressing concern for media owners, says Narayanan Rajan, CEO of Media Excel. “They are trying to save costs. As more content owners are streaming directly to consumers, cloud egress and CDN costs start to dominate Total Cost of Ownership considerations. It is becoming even more critical to reduce the cost of scaling streamed video.”
The 20% efficiency gain is enabled thanks to real-time analysis of incoming video streams with dynamic adjustment of the encoding settings – all resulting in optimized codec performance. Rajan points out that a generational shift to a new codec requires adoption across the delivery ecosystem including on edge devices, which is beyond the control of media owners. He calls that an “expensive uplift” that DIVA can help to avoid for some time.
Indeed, his company is looking to push the AI-enabled efficiency gains towards 30%, although Rajan will only commit to this as a roadmap item right now. “We see 30% as realistic but there is still work to do to figure out how we get there.”
Media Excel is claiming “unparallelled visual quality” for a solution that works with both constant and variable bit rate delivery. DIVA encoding will be enabled via on-premise Media Excel Hero appliances or in containerized instances in public clouds.
The company is also introducing a suite of DIVA pre-processing features later this year such as super-resolution, image restoration and automated subtitle generation in multiple languages. The subtitling uses AI for audio detection and will be designed for live television, meaning the content has to be assessed, then translated multiple times (potentially) in parallel, with the resulting live output not suffering from any noticeable latency.
Indeed, all these advances are being architected with an eye on latency and at IBC Media Excel will be highlighting its super-fast transcoding and packaging engine. Rajan says the Media Excel video pipeline contributes significantly to efforts to get streaming latency to broadcast levels.
