Filling the Gaps in Video Transcoder Deployment in the Cloud
Originally Aired - Monday, April 17 | 10:00 AM - 10:20 AM PT
This Session Has Not Started Yet
Be sure to come back after the session starts to have access to session resources.
Cloud based deployment of content production and broadcast workflows has continued to disrupt the industry after the pandemic. However, optimising the deployment of transcoding infrastructure is still a challenge. Firstly, per-clip content adaptation provides huge gains in compression but comes with substantial compute cost. Secondly, pre-processing of media is well known to improve compression behaviour, but the control of that pre-processor is difficult. Thirdly, users cannot yet with confidence assess or predict the costs of these modules in the cloud when applied to their datasets without launching the job and making a crude prediction as the job unfolds.
In this talk we present several low-cost optimisation strategies for filling the gaps between the realities of cloud transcoder deployment today and the potential gains that these systems can offer. We show an empirical pre-processor optimisation technique, evolution of per-clip optimisation with proxy resolution. Lastly, we showcase mechanism to predict transcoding time with high accuracy with low bandwidth features for a usable cloud cost calculator that is missing from the current arsenal of tools.