GPU-Accelerated FFV1 Encoding Is Here - a Case Study Using Temporal for Durable Execution

GPU-based video encoding is widely available, efficient, and performant, but it has not been adopted in Digital Preservation tooling due to a lack of support for preservation codecs and a quality deficit when working with more common consumer codecs. Recently it has become possible to encode preservation-standard FFV1 video in ffmpeg using a GPU, which effectively addresses all previously-existing limitations of this approach. This short paper frames the current state of the art as well as why this development is so potentially useful, and provides a case study of using GPU-accelerated FFV1 encoding with Temporal, a Durable Execution platform that can enable further scaling.