Post-Compression Quality Check: Catching Overtightened Encodes
The trap of compression is not failing to compress — it is compressing too hard and not noticing until viewers see mush. A four-minute self-check stands between your encode and your audience.
Three damage types and how to spot them
Blocking artifacts: mosaic grids or banded steps in flat areas — sky, walls, solid backgrounds. Starved encoders sacrifice flat regions first; look there. Detail mush: dense textures (foliage, hair, text edges) turn smeary, worse during motion.
Motion smearing and banding: fast movement leaves trailing ghosts; dark gradients (night scenes, dusk) break into contour stripes. The "dark + moving + textured" shot is where all three erupt together — it is always your first stop.
The spot-check: three fixed moves
First, jump to the darkest moving shot and watch 10 seconds (low light amplifies every artifact). Second, pause on any text and inspect its edges — sharpness of letterforms is the most sensitive quality indicator. Third, play at full speed once and feel for stutter — bitrate starvation shows as both stutter and artifacts. Three moves, three minutes, far better than "watching it through".
The stricter version: put both versions side by side in one frame with a comparison tool, playing in sync — artifacts and mush become instantly obvious. Worth it for client deliveries and doubtful cases.
Backing off and preventing the crush
Found it crushed: go back to the source and re-encode lighter — each 6-point CRF reduction roughly doubles the bitrate; stepping 30-to-26 or 26-to-23 are common retreats. Never "lightly re-compress" the compressed file — that stacks a second lossy generation on top.
Prevention numbers: CRF 26 to start, 23 for delivery-grade, 20 for near-transparent. And test-first: encode a 30-second hardest-moment excerpt, run the three moves on it, then batch the full video — a one-minute pre-flight replaces a full re-encode.