A dependency-light Python script (standard library only) that reads a libvmaf JSON log — with or without the bootstrap confidence-interval fields — and reports the score pooled four ways (mean, harmonic mean, 5th percentile, single worst frame), prints the 95% confidence interval when a bootstrap model was used (both the percentile [ci95_low, ci95_high] and the score ± 1.96·stddev normal forms), and, given two logs, returns a significance verdict for the gap read against both the overlapping err
Download free PDF
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
This analyzer reads a libvmaf JSON log, pools the score four ways, prints the 95% VMAF confidence interval and tells you whether the gap between two encodes is significant. Our seven-volume handbook covers the methods: Volume 1 on VMAF models and quality metrics, Volume 3 on ladders and QoE, and Volume 6 on the ML basics behind models. Written by Nikolay Sapunov, CEO at Fora Soft.
Is the VMAF gap between two encoders real or just noise?
Send us your two encodes and logs: we will run the significance check, explain the result and recommend which encoder settings to keep. 250+ video products shipped since 2005.
You get a realistic estimate, technical recommendations and next steps. Free, no obligations.
