From b9d00885981c0532b8730c4feae087f91b51a375 Mon Sep 17 00:00:00 2001 From: gravity-zero Date: Wed, 18 Nov 2026 14:05:02 +0100 Subject: [PATCH] docs: grid-timed audio window no longer experimental, player measurements stated --- CHANGELOG.md | 8 ++++++++ docs/library.md | 8 ++++++-- 2 files changed, 14 insertions(+), 2 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 5d4ce3f..fa35adf 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -9,6 +9,14 @@ its series' file in [changelog/](changelog/) when a newer one ships. ## [Unreleased] +### Changed + +- Grid-timed audio whose timecodes run a frame ahead of the grid for a while + is no longer described as experimental: measured in hls.js, Shaka and + ExoPlayer, no player stalls on the window that absorbs the frame; Firefox + keeps the audio up to about a frame late past it, as before 0.62.0 + (`docs/library.md`, HLS plan). + ## [0.69.0] - 2026-11-17 ### Added diff --git a/docs/library.md b/docs/library.md index 7d87081..7a82b1d 100644 --- a/docs/library.md +++ b/docs/library.md @@ -319,8 +319,12 @@ first block's slot (no earlier than where the window before ends, when its timecodes leave the grid), and ends where the next one opens. A run of stored timecodes a frame ahead of the grid leaves the frames after it a frame late; the window where the run ends gives that frame back, its frames from there -slightly shorter, and the audio after it is on its own timecodes again. This -rule is experimental: how players render that window is still being measured. +slightly shorter, and the audio after it is on its own timecodes again; a +segment served alone carries the same bytes whatever order the segments are +asked in. Measured on a run of +32 ms: hls.js and Shaka in Chrome, and +ExoPlayer, play the samples back to back and stay in sync with or without it; +Firefox follows the timestamps and keeps the audio up to about a frame late +past the window, as it did before this rule. Safari is not measured. `HLSPlan.CueDensity()` reports what the plan cut on - the number of video cue points, the widest gap between two of them and where it opens, the median gap, the last one, and whether `SynthesizeIndex` built them (then every