Generative AI Uses and Video Transcoding

Speech Intelligibility and Audio Monitoring in OTT - $15

The paper details ways of controlling speech intelligibility in OTT and broadcast production, primarily based on a new metric, Loudness to Dialog Ratio (LDR), and on calibrated sound monitoring at a moderate listening level. Principles of Loudness normalization are described; as is a study to define and quantify LDR for use in broadcast and OTT. Findings may be applied during mixing or dubbing, used for default settings in NGA delivery, and in automated QC as a key machine learning parameter. The article is facts-based and free of commercial bias.

Thomas Lund | Genelec Inc | Natick, Mass., United States

The Power of Generative AI for Personalizing Video Content - $15

In today’s rapidly evolving landscape of content creation, a brave new world is emerging; video service providers can utilize AI-assisted pipelines to streamline the content creation process. As technology advances, large language model (LLM)-powered tools are ushering in a new era for content personalization and targeting. This paper presents an architecture blueprint to boost the effectiveness of targeted ads by developing tailored messages for specific consumer segments focusing on their core values and beliefs. Central to this are three pillars: automated audience segmentation and identifying deep seated core values, GenAI-aided message crafting aligned with segment values and beliefs, and a closed-loop system for continuous improvement through automated A/B testing. This paper calls for a holistic approach to video personalization that considers the enforcement of ethical — particularly privacy-related — standards and appropriate advertisers’ guardrails, as well as the prevention of biases. By leveraging data-driven insights and continuous feedback, we aim to refine marketing strategies, ensuring messages effectively resonate with targeted groups.

Ofer Weintraub | Viaccess-Orca | Raanana, Israel
Alice Wittenberg | Viaccess-Orca | Raanana, Israel
Alain Nochimowski | Viaccess-Orca | Raanana, Israel

Unravelling the Power of Single-Pass Look-Ahead in Modern Codecs for optimised transcoding deployment - $15

Winner of the 2024 NAB BEIT Conference Proceedings Best Student Paper Award

Modern video encoders have evolved into sophisticated pieces of software in which various coding tools interact with each other. In the past, single-pass encoding was not considered for Video-On-Demand (VOD) use-cases. In this work, we evaluate production-ready encoders for H.264 (x264), H.265 (HEVC), AV1 (SVT-AV1) along with direct comparisons to the latest AV1 encoder inside NVIDIA GPUs (40 series), and AWS Mediaconvert’s AV1 implementation. Our experimental results demonstrate single pass encoding inside modern encoder implementations can give us very good quality at reasonable compute cost. The results are presented as three different scenarios targeting High, Medium, Low complexity accounting quality-bitrate-compute load. Finally, a set of recommendations are presented for end-users to help decide which encoder/preset combination might be more suited to their use case.

Vibhoothi | Sigmedia Group, Department of Electrical Engineering | Trinity College Dublin, Ireland
Julien Zouein | Sigmedia Group, Department of Electrical Engineering | Trinity College Dublin, Ireland
François Pitié | Sigmedia Group, Department of Electrical Engineering | Trinity College Dublin, Ireland
Anil Kokaram | Sigmedia Group, Department of Electrical Engineering | Trinity College Dublin, Ireland