Introduce Dahua CGI source for 2-way audio

.gitignore

@@ -1,6 +1,7 @@
 .idea/
 .tmp/
 
+go2rtc
 go2rtc.yaml
 go2rtc.json
 
@@ -12,3 +13,7 @@ go2rtc_win*
 0_test.go
 
 .DS_Store
+
+*.mp3
+*.wav
+*.al

fix_vto_codecs.sh

@@ -0,0 +1,103 @@
+#!/bin/bash
+#
+# This script fixes the VTO audio codecs before supplying the stream url to go2rtc.
+#
+# Usage: ./fix_vto_codecs.sh [--debug] [--https] <VTO stream URL>
+#
+# Examples:
+#
+#   ./fix_vto_codecs.sh rtsp://user:[email protected]/cam/realmonitor?channel=1&subtype=0&unicast=true&proto=Onvif
+#   ./fix_vto_codecs.sh rtsp://user:[email protected]/cam/realmonitor?channel=1&subtype=1
+#
+#   If the VTO has HTTPS enabled:
+#   ./fix_vto_codecs.sh --https rtsp://user:[email protected]/cam/realmonitor?channel=1&subtype=0
+#
+#   With ffmpeg prefix:
+#   ./fix_vto_codecs.sh ffmpeg:rtsp://user:[email protected]/cam/realmonitor?channel=1&subtype=0#video=copy#audio=copy
+#
+
+set -euo pipefail
+
+usage() {
+  echo "Usage: ${0} [--debug] [--https] <VTO stream URL>" >&2
+  exit "${1}"
+}
+
+debug=false
+protocol="http"
+extra_curl_args=()
+
+positional_args=()
+while [[ $# -gt 0 ]]; do
+  case $1 in
+    --debug)
+      debug=true
+      shift
+      ;;
+    --https)
+      protocol="https"
+      extra_curl_args+=("--insecure")
+      shift
+      ;;
+    -*)
+      echo "Unknown option ${1}" >&2
+      usage 1
+      ;;
+    *)
+      positional_args+=("$1")
+      shift
+      ;;
+  esac
+done
+
+set -- "${positional_args[@]}"
+unset positional_args
+
+if [[ $# -ne 1 ]]; then
+  echo "Expected 1 positional argument, got $#" >&2
+  usage 1
+fi
+
+vto_stream_url="${1}"
+
+if [[ "${vto_stream_url}" != "rtsp://"* && "${vto_stream_url}" != "ffmpeg:rtsp://"* ]]; then
+  echo "VTO stream URL does not start with rtsp:// or ffmpeg:rtsp://" >&2
+  usage 1
+fi
+
+if [[ "${debug}" == "true" ]]; then
+  set -x
+  extra_curl_args+=("--verbose")
+fi
+
+vto_host_with_creds="${vto_stream_url#"ffmpeg:"}"
+vto_host_with_creds="${vto_host_with_creds#"rtsp://"}"
+vto_host_with_creds="${vto_host_with_creds%%"/"*}"
+if [[ "${vto_host_with_creds}" =~ :[0-9]+$ ]]; then
+  vto_host_with_creds="${vto_host_with_creds%":"*}"
+fi
+
+query="action=setConfig"
+# PCMA: average audio quality, but good for WebRTC and 2-way audio
+query+="&Encode[0].MainFormat[0].Audio.Compression=G.711A"
+# 16000Hz yields better audio quality, but try 8000Hz if your VTO does not support it
+query+="&Encode[0].MainFormat[0].Audio.Frequency=8000"
+# AAC: best audio quality, good for Frigate recordings
+query+="&Encode[0].ExtraFormat[0].Audio.Compression=AAC"
+
+# PS: the current config can be retrieved with:
+#   curl -fsSL --digest --globoff \
+#     http://user:[email protected]/cgi-bin/configManager.cgi?action=getConfig&name=Encode
+
+output=$(
+  curl --fail --silent --show-error --digest --globoff "${extra_curl_args[@]}" \
+    "${protocol}://${vto_host_with_creds}/cgi-bin/configManager.cgi?${query}"
+)
+
+if [[ "${output}" != $'OK\r' ]]; then
+  echo "Failed to set VTO codecs. Response:" >&2
+  echo "${output}" >&2
+  exit 1
+fi
+
+echo -n "${vto_stream_url}"

internal/dahua/init.go

@@ -0,0 +1,13 @@
+package dahua
+
+import (
+	"github.com/AlexxIT/go2rtc/internal/streams"
+	"github.com/AlexxIT/go2rtc/pkg/core"
+	"github.com/AlexxIT/go2rtc/pkg/dahua"
+)
+
+func Init() {
+	streams.HandleFunc("dahua", func(source string) (core.Producer, error) {
+		return dahua.Dial(source)
+	})
+}

main.go

@@ -6,6 +6,7 @@ import (
 	"github.com/AlexxIT/go2rtc/internal/api/ws"
 	"github.com/AlexxIT/go2rtc/internal/app"
 	"github.com/AlexxIT/go2rtc/internal/bubble"
+	"github.com/AlexxIT/go2rtc/internal/dahua"
 	"github.com/AlexxIT/go2rtc/internal/debug"
 	"github.com/AlexxIT/go2rtc/internal/doorbird"
 	"github.com/AlexxIT/go2rtc/internal/dvrip"
@@ -84,6 +85,7 @@ func main() {
 	dvrip.Init()    // dvrip source
 	tapo.Init()     // tapo source
 	isapi.Init()    // isapi source
+	dahua.Init()    // dahua source
 	mpegts.Init()   // mpegts passive source
 	roborock.Init() // roborock source
 	homekit.Init()  // homekit source

pkg/dahua/CGI_SDK_API.txt

@@ -0,0 +1,154 @@
+11. Audio
+
+11.1 Audio MIME type
+
+MIME                                                                   Description
+Audio/PCM
+Audio/ADPCM
+Audio/G.711A
+Audio/G.711Mu
+Audio/G.726
+Audio/G.729
+
+                                                                  92
+Audio/MPEG2
+Audio/AMR
+Audio/AAC
+
+
+
+11.2 Post Audio
+
+URL Syntax       http://<ip>/cgi-bin/audio.cgi?action=postAudio&<paramName>=<paramValue>[&<paramName>=<paramValue>...]
+Comment          paramValue as below table.
+Response         OK or ERROR
+
+
+ParamName                                         ParamValue type                   Description
+httptype                                          string                            singlepart:HTTP content is a continuous flow of audio
+                                                                                    packets
+                                                                                    multipart:HTTP           content      type         is
+                                                                                    multipart/x-mixed-replace,and each audio packet ends
+                                                                                    with a boundary string
+channel                                           integer                           The audio channel
+
+
+
+11.2.1 Example for singlepart
+
+The RUL of transmit a singlepart、channel 1 audio stream(encoded with G.711 A-law) is:
+http: //<ip>/cgi-bin/audio.cgi?action=postAudio&httptype=singlepart&channel=1
+
+
+example:
+POST /cgi-bin/audio.cgi?action=postAudio&httptype=singlepart&channel=1 HTTP/1.1
+Content-Type: Audio/G.711A
+Content-Length:9999999
+
+
+<Audio data>
+<Audio data>
+
+
+
+11.2.2 Example for multipart
+
+The RUL of transmit a multipart、channel 1 audio stream(encoded with G.711 A-law) is:
+http: //<ip>/cgi-bin/audio.cgi?action=postAudio&httptype= multipart &channel=1
+
+
+example:
+POST /cgi-bin/audio.cgi?action=postAudio&httptype= multipart &channel=1 HTTP/1.1
+ Content-Type: multipart/x-mixed-replace; boundary=<boundary>
+--<boundary>
+Content-Type: Audio/G.711A
+
+                                                                    93
+Content-Length: 800
+
+
+<Audio data>
+--<boundary>
+
+
+
+11.3 Get Audio
+
+URL Syntax       http://<ip>/cgi-bin/audio.cgi?action=getAudio&<paramName>=<paramValue>[&<paramName>=<paramValue>...]
+Comment          paramValue as below table.
+Response         OK or ERROR
+
+
+
+
+ParamName                                           ParamValue type                    Description
+httptype                                            string                             singlepart:HTTP content is a continuous flow of audio
+                                                                                       packets
+                                                                                       multipart:HTTP           content        type         is
+                                                                                       multipart/x-mixed-replace,and each audio packet ends
+                                                                                       with a boundary string
+channel                                             integer                            The audio channel
+
+
+
+
+11.3.1 Example for singlepart
+
+The RUL of Request a singlepart、channel 1 audio stream(encoded with G.711 A-law) is:
+http: //<ip>/cgi-bin/audio.cgi?action=getAudio&httptype=singlepart&channel=1
+
+
+If the request was successful, the server returns a continuous flow of audio packets.The content type is only set at the beginning of the
+connection.
+Return:
+HTTP Code: 200 OK
+Content-Type: Audio/G.711A
+Body:
+<Audio data>
+<Audio data>
+
+
+
+11.3.2 Example for multipart
+
+The RUL of Request a multipart、channel 1 audio stream(encoded with G.711 A-law) is:
+http: //<ip>/cgi-bin/audio.cgi?action=getAudio&httptype=multipart&channel=1
+
+
+If the request was successful, the server returns a continuous flow of audio packets. The content type is “multipart/x-mixed-replace” and each
+audio packet ends with a boundary string.
+                                                                      94
+Return:
+HTTP Code: 200 OK
+Content-Type: multipart/x-mixed-replace; boundary=<boundary>
+--<boundary>
+Content-Type: Audio/G.711A
+Content-Length: 800
+
+
+<Audio data>
+--<boundary>
+
+
+
+
+11.4 Audio Input
+
+11.4.1 getCollect
+
+URL Syntax            http://<ip>/cgi-bin/devAudioInput.cgi?action=getCollect
+Comment               Get Audio input channel number.
+                      Below response means there are 2 audio input channels.
+Response              result=2
+
+
+
+
+11.5 Audio Output
+
+11.5.1 getCollect
+
+URL Syntax            http://<ip>/cgi-bin/devAudioOutput.cgi?action=getCollect
+Comment               Get Audio output channel number.
+                      Below response means there are 2 audio output channels.
+Response              result=2