From 55baf868e6982f61e3c52a8cc5c19fabfe31fda1 Mon Sep 17 00:00:00 2001
From: Eli Mallon <eli@stream.place>
Date: Fri, 13 Jun 2025 12:57:39 -0700
Subject: [PATCH] app: opusparse on ingest, seems to fix a/v duration mismatch
 (#286)

* app: opusparse on ingest, seems to fix a/v duration mismatch

* hack: add segment-durations.sh
---
 hack/segment-durations.sh  | 16 ++++++++++++++++
 pkg/media/webrtc_ingest.go |  2 +-
 2 files changed, 17 insertions(+), 1 deletion(-)
 create mode 100755 hack/segment-durations.sh

diff --git a/hack/segment-durations.sh b/hack/segment-durations.sh
new file mode 100755
index 00000000..491e0b9a
--- /dev/null
+++ b/hack/segment-durations.sh
@@ -0,0 +1,16 @@
+#!/bin/bash
+
+# Splits a video stream apart into video and audio files and checks the discrepancy between them
+
+set -euo pipefail
+
+dir="$(mktemp -d)"
+ffmpeg -i "$1" -vn -c:a copy "$dir/audio.mkv" 2>/dev/null >/dev/null
+ffmpeg -i "$1" -an -c:v copy "$dir/video.mkv" 2>/dev/null >/dev/null
+videoDuration=$(gst-discoverer-1.0 "$dir/video.mkv" | grep "Duration" | sed 's/  Duration: 0:00:0//')
+audioDuration=$(gst-discoverer-1.0 "$dir/audio.mkv" | grep "Duration" | sed 's/  Duration: 0:00:0//')
+
+echo "Video duration: $videoDuration"
+echo "Audio duration: $audioDuration"
+echo "Difference (negative means audio is longer): $(node -p "$videoDuration - $audioDuration")"
+rm -rf "$dir"
diff --git a/pkg/media/webrtc_ingest.go b/pkg/media/webrtc_ingest.go
index 200bfbc1..4b5e2b62 100644
--- a/pkg/media/webrtc_ingest.go
+++ b/pkg/media/webrtc_ingest.go
@@ -34,7 +34,7 @@ func (mm *MediaManager) WebRTCIngest(ctx context.Context, offer *webrtc.SessionD
 	pipelineSlice := []string{
 		"multiqueue name=queue",
 		"appsrc format=time is-live=true do-timestamp=true name=videosrc ! capsfilter caps=application/x-rtp ! rtph264depay ! capsfilter caps=video/x-h264,stream-format=byte-stream,alignment=nal ! h264parse disable-passthrough=true config-interval=-1 ! h264timestamper ! identity ! queue.sink_0",
-		"appsrc format=time do-timestamp=true name=audiosrc ! capsfilter caps=application/x-rtp,media=audio,encoding-name=OPUS,payload=111 ! rtpopusdepay ! queue.sink_1",
+		"appsrc format=time do-timestamp=true name=audiosrc ! capsfilter caps=application/x-rtp,media=audio,encoding-name=OPUS,payload=111 ! rtpopusdepay ! opusparse ! queue.sink_1",
 	}
 
 	pipeline, err := gst.NewPipelineFromString(strings.Join(pipelineSlice, "\n"))
-- 
2.51.2