From 18b32431b442285adeaf0f43e27a1ba5975e92f2 Mon Sep 17 00:00:00 2001 From: Chris Cannam Date: Thu, 23 Mar 2023 17:49:59 +0000 Subject: [PATCH] Update changelog --- CHANGELOG | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) diff --git a/CHANGELOG b/CHANGELOG index 7edf8bc..dc5a8fe 100644 --- a/CHANGELOG +++ b/CHANGELOG @@ -1,9 +1,16 @@ Changes in Rubber Band v3.2 -This release contains a number of fixes, including some which affect -the output quality in certain processing modes. +This significant release contains a number of fixes, including some +which affect the output quality in certain processing modes. + * Fix inability of the R3 engine to produce mono-compatible output + when processing stereo. In R2 the OptionChannelsTogether option has + always done this (with some quality caveats). In R3 this option was + more nearly mono-compatible than the default, but not + entirely. With this release, OptionChannelsTogether uses a new + method that is both high quality and mono-compatible and is + recommended for situations benefiting from this. * Correct the start delay calculation in "resample-before" modes of the R3 stretcher. This affects latency in realtime mode when using the non-recommended realtime pitch options (i.e. those other than @@ -14,6 +21,9 @@ the output quality in certain processing modes. * Fix failure to generate output when called with final=true and no samples at input (a regression in v3.1, this worked correctly in v3.0 and in the R2 engine) + * Fix small difference in output from R3 engine depending on whether + time and pitch ratios were provided on construction or immediately + after it * Fix occasional truncation of final block when using Speex resampler * Fix hang in the command-line utility code at end of processing with certain file formats