Fix A/V desync when resuming HLS with video transcode + audio copy (#16580)

Fix A/V desync when resuming HLS with video transcode + audio copy

MediaBrowser.Controller/MediaEncoding/EncodingHelper.cs

@@ -86,6 +86,7 @@ namespace MediaBrowser.Controller.MediaEncoding
         private readonly Version _minFFmpegQsvVppScaleModeOption = new Version(6, 0);
         private readonly Version _minFFmpegRkmppHevcDecDoviRpu = new Version(7, 1, 1);
         private readonly Version _minFFmpegReadrateCatchupOption = new Version(8, 0);
+        private readonly Version _minFFmpegNoiseBsfDrop = new Version(5, 0);
 
         private static readonly string[] _videoProfilesH264 =
         [
@@ -1547,20 +1548,61 @@ namespace MediaBrowser.Controller.MediaEncoding
 
         public string GetAudioBitStreamArguments(EncodingJobInfo state, string segmentContainer, string mediaSourceContainer)
         {
-            var bitStreamArgs = string.Empty;
+            var filters = new List<string>();
+
+            var noiseFilter = GetCopiedAudioTrimBsf(state);
+            if (!string.IsNullOrEmpty(noiseFilter))
+            {
+                filters.Add(noiseFilter);
+            }
+
             var segmentFormat = GetSegmentFileExtension(segmentContainer).TrimStart('.');
 
             // Apply aac_adtstoasc bitstream filter when media source is in mpegts.
             if (string.Equals(segmentFormat, "mp4", StringComparison.OrdinalIgnoreCase)
                 && (string.Equals(mediaSourceContainer, "ts", StringComparison.OrdinalIgnoreCase)
                     || string.Equals(mediaSourceContainer, "aac", StringComparison.OrdinalIgnoreCase)
-                    || string.Equals(mediaSourceContainer, "hls", StringComparison.OrdinalIgnoreCase)))
+                    || string.Equals(mediaSourceContainer, "hls", StringComparison.OrdinalIgnoreCase))
+                && IsAAC(state.AudioStream))
             {
-                bitStreamArgs = GetBitStreamArgs(state, MediaStreamType.Audio);
-                bitStreamArgs = string.IsNullOrEmpty(bitStreamArgs) ? string.Empty : " " + bitStreamArgs;
+                filters.Add("aac_adtstoasc");
             }
 
-            return bitStreamArgs;
+            return filters.Count == 0
+                ? string.Empty
+                : " -bsf:a " + string.Join(',', filters);
+        }
+
+        // When video is transcoded, accurate_seek (the default) trims video to the
+        // exact seek point via decoder-side frame discard. But stream-copied audio
+        // bypasses the decoder, so it starts from the nearest keyframe — potentially
+        // seconds before the target. Use the noise bsf to drop copied audio packets
+        // before the seek target, achieving the same trim precision without
+        // re-encoding. The noise bsf's drop= parameter requires ffmpeg >= 5.0.
+        // Important: make sure not to use it with wtv because it breaks seeking
+        private string GetCopiedAudioTrimBsf(EncodingJobInfo state)
+        {
+            if (state.TranscodingType is not TranscodingJobType.Hls
+                || !state.IsVideoRequest
+                || IsCopyCodec(state.OutputVideoCodec)
+                || !IsCopyCodec(state.OutputAudioCodec)
+                || string.Equals(state.InputContainer, "wtv", StringComparison.OrdinalIgnoreCase)
+                || _mediaEncoder.EncoderVersion < _minFFmpegNoiseBsfDrop)
+            {
+                return null;
+            }
+
+            var startTicks = state.BaseRequest.StartTimeTicks ?? 0;
+            if (startTicks <= 0)
+            {
+                return null;
+            }
+
+            var seekSeconds = startTicks / (double)TimeSpan.TicksPerSecond;
+            return string.Format(
+                CultureInfo.InvariantCulture,
+                "noise=drop='lt(pts*tb\\,{0:F3})'",
+                seekSeconds);
         }
 
         public static string GetSegmentFileExtension(string segmentContainer)
@@ -3006,23 +3048,6 @@ namespace MediaBrowser.Controller.MediaEncoding
                 }
 
                 seekParam += string.Format(CultureInfo.InvariantCulture, "-ss {0}", _mediaEncoder.GetTimeParameter(seekTick));
-
-                if (state.IsVideoRequest)
-                {
-                    // If we are remuxing, then the copied stream cannot be seeked accurately (it will seek to the nearest
-                    // keyframe). If we are using fMP4, then force all other streams to use the same inaccurate seeking to
-                    // avoid A/V sync issues which cause playback issues on some devices.
-                    // When remuxing video, the segment start times correspond to key frames in the source stream, so this
-                    // option shouldn't change the seeked point that much.
-                    // Important: make sure not to use it with wtv because it breaks seeking
-                    if (state.TranscodingType is TranscodingJobType.Hls
-                        && string.Equals(segmentContainer, "mp4", StringComparison.OrdinalIgnoreCase)
-                        && (IsCopyCodec(state.OutputVideoCodec) || IsCopyCodec(state.OutputAudioCodec))
-                        && !string.Equals(state.InputContainer, "wtv", StringComparison.OrdinalIgnoreCase))
-                    {
-                        seekParam += " -noaccurate_seek";
-                    }
-                }
             }
 
             return seekParam;