Systematic Benchmarking of 360° Virtual Reality Video Production: A Multi-Experiment Study of Encoding, Audio Systems, and Workflow

Abstract

Producing 360° virtual reality (VR) documentaries requires evidence about processing costs and deployment constraints. This study benchmarks five production stages using material from Beneath the Mist. Original measurement workbooks provide the primary numerical evidence; later reconstructions support separate verification. Across 30 resolution records, mean stitching time increases from 1.502 minutes at 4K to 1.843 minutes at 6K and 2.307 minutes at 8K, indicating sub-linear growth with respect to pixel count, without establishing a hardware mechanism. Fifteen audio timing blocks show no detected presentation, layout, or interaction effect in the common VR/two-dimensional (2D) stereo/5.1 comparison. The first-order ambisonics (FOA) and legacy extended-audio columns take longer. Later audio-only reconstructions show identical duration-matched stereo-to-5.1 byte increments in VR and 2D; the original total video-file-size hypothesis remains directly untested. Across 20 workflow records, nadir cleanup has the longest duration under a provisional manual-time convention. Two headset-transfer workflows show different recorded intervals, but their connector and interface paths are conflated, and physical throughput has not been verified. Across five segments and 120 encoding outputs, the preset-label timing effect survives a two-record sensitivity check, whereas no Video Multimethod Assessment Fusion (VMAF) effect survives multiplicity adjustment. The companion workbook supports the primary and two-record analyses; reproducing the archived exact-size and full-timestamp results requires their complete row-level records. These configuration-specific benchmarks inform production planning and replication within the stated measurement scope.