Block 5. Production QC

Automated video quality control means gates that pass, warn, or block. The three QC checkpoints, the EBU layer model, and hard vs soft gates explained.

Per-title and per-shot encoding explained from the quality side: how a VMAF target, not a fixed bitrate, defines every rung of an adaptive bitrate ladder.

Convex hull encoding explained: why resolution curves cross, how to build the hull without corrupting it, and how it drives per-title ladders.

Set a VMAF target instead of a bitrate: where quality goes invisible (~95), per-platform targets, and turning the target into a delivery budget.

Build a video quality gate in CI/CD: gate on VMAF not PSNR, run a floor plus a golden-reference regression check, size the margin from metric noise.

Build a video quality regression suite: golden reference clips, a noise-sized tolerance band not an exact line, and a frozen baseline to catch slow drift.

Monitor delivered video quality after launch: no-reference bitstream models, player telemetry, sampling at scale, and per-segment alerting on drops.

Build a video quality report a stakeholder trusts: report the worst frames, show confidence intervals, draw honest charts, and attach provenance.