Freezing - more precisely a stall or rebuffering - is the event where playback stops because the player's buffer ran dry, and the last decoded frame sits frozen on screen (often under a spinner) until enough data arrives to resume. Over reliable transport like the TCP or QUIC under HLS and DASH, a network problem does not corrupt pixels; lost packets are retransmitted, which takes time, so the buffer drains and you get a freeze. It is the artifact that defeats a picture metric most completely, because during a freeze every frame on screen is a correct, fully decoded frame - simply the same correct frame held too long, so a full-reference PSNR, SSIM, or VMAF finds nothing wrong. The fault is in time, and a per-frame score has no time axis. Freezing also carries the hardest business numbers: a rebuffering ratio equal to 1% of duration costs about 5% of watch time, and startup past 2 seconds raises abandonment by roughly 5.8% per added second. Measure it with player-side QoE metrics (CTA-2066) and session models (ITU-T P.1203/P.1204), not picture scores.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
Freezing is invisible to picture metrics because every frame is correct, so it has to be measured from the player's timeline. We wrote seven volumes covering the whole video stack, and for freezing start with Volume 3: buffers and stall recovery, ABR, CDNs and QoE monitoring, with Volume 2 on HLS and DASH delivery. Written by Nikolay Sapunov, CEO at Fora Soft.
Viewers complaining about freezing while your quality dashboards look green?
Share your player and CDN data: we locate where freezing starts and give you buffer, ABR and delivery changes to reduce rebuffering. 250+ video products shipped since 2005.
