Support for audio recording

[kankadev]

Hello,

I use motion as surveillance system and I'm missing the possibility to record audio with an USB microphone or with an USB webcam with builtin mic, too.

I tried to code something to start audio recording manually when the event on_motion_detected is triggered, but it's difficult to stop the audio recording when motion is over and to compare video and audio into one file.

[af7567]

Has anyone done any work on audio recording built into motion, rather than using on_motion_detected and _ended scripts?

I had a quick look at the code the other day thinking I might spot somewhere I can just hack it in quickly to try it out, but soon realised it's not that easy to pass the input data to the output data. :)

I am more interested in recording from an RTSP stream than from USB, and while it can be done with scripts it seems wrong to open another RTSP connection to the camera just to grab the audio since that data is already being sent to motion. It also takes a few seconds for ffmpeg to be able to start capturing audio from a new connection even on the substream, so it would be better if it can be done from the already open stream.

[Mr-DaveDev]

The only "work" is within the MotionPlus application but it is not working very well. There is an extreme latency. I would never be putting this into Motion so you'd have to wait for the next maintainer if you want it inside of Motion.

Everyone should also be acutely aware of the legal ramifications. In "most" countries it is fine to record video but illegal to record audio security footage of people. (To my knowledge, "most" includes: USA, Canada, Europe, Australia, NZ and probably many more). It would be real bummer to submit the recording to the authorities for something happening to your property and then you are the one ending up with a felony for wiretapping!

[pfzim]

Why do you rely only on US laws? Sound recording may not be prohibited in other countries. All commercial products can record sound from cameras.

Finally, make audio recording optional. Add sound recording at least from RTSP stream, I think it will not be difficult for you.

[gene-git]

My comment was obviously about US laws only. I am not familiar with those in other countries and Dave is the dev not me.
I'm just a happy user :)

That said, maybe you should try MotionPlus. As @Mr-Dave said early on in this thread, work on sound has been happening there.

I can say I switched over from Motion to MotionPlus and audio is working just fine for me with rtsp streams - or at least as fine as the camera mic provides :)

[af7567]

I don't know why I never looked at MotionPlus before, I should give that a go :)
In the UK it is legal to record audio and video in your driveway/garden for peronal use. The only real problems would be if you share those recordings with third parties without getting permission from the people who were recorded.

[Mr-Dave]

Please go elsewhere.

[gene-git]

I also note that connecting to rtsp stream using VLC i get video and audio - however going through motion there is video only.
Obviously with multiple cameras on the main web page audio should be off for all but one (or none) - but when watching an individual camera it would be nice if audio was on - this is all audio + video over network no usb obviously.

Is there something I'm have not set up correctly I wonder?

[gene-git]

Following on with audio vs video re cording - in the US video is generally fine, but its considered polite to inform. Audio rules/laws vary by state - some states are 2 party consent while some are 1 party - but even in the latter case, there must be 1 party - and if "you" are not present then its not okay. Once again, the standard is to inform that recording is happening. There are things applicable to phone calls (periodic beeps for example) that do not make sense for surveillance - so one need ensure that all parties are made aware - then it's fine.

[meirlazar]

I've created a hacky solution for recording audio as a separate process, and it syncs up very well with the video
The motionaudio script will be called in the motion.conf file with 2 arguments (either start or stop the audio recording) and the 2nd argument uses the same format used for the movie_filename. This will generate a file.wav and file.pid while the file.avi (video file) is created by motion. The stop argument kills the arecord process and combines the audio and video file into a new file and removes the other 2 files.

motion.conf

ffmpeg_video_codec mpeg4 # produces .avi files
movie_filename %m-%d-%Y_%H:%M:%S
on_event_start /bin/bash /path/to/motionaudio.sh start %m-%d-%Y_%H:%M:%S
on_event_end  /bin/bash /path/to/motionaudio.sh stop %m-%d-%Y_%H:%M:%S

motionaudio.sh

#!/bin/bash
moviedir=/path/to/motion/videos # change this 
action=$1
audio=$2
aext=wav # change if you prefer a different audio format or extension
vext=avi # change this to what is specified in the motion.conf file for the line ffmpeg_video_codec
voutext=mkv #change this to the video file that will be created by joining the audio and video files

function startaudio () {
  if pgrep -f "arecord"; then sudo pkill -9 -f "arecord -f cd -r 22050 -D plughw:2,0"; fi
  arecord -f cd -r 22050 -D plughw:2,0 "${moviedir}/${audio}.${aext}" &
  audiopid=${BASHPID}
  echo "${audiopid}" > "${moviedir}/${audio}.pid"
}

function stopaudio () {
  audiopids=$(find ${moviedir} -type f -iname "*.pid" -printf "%f\n" | cut -d. -f1)
  while read -r x; do 
	  if pgrep -F "${x}.pid"; then sudo pkill -9 -F "${x}.pid"; fi
	  if [[ -f "${moviedir}/${x}.${vext}" ]] && [[ -f "${moviedir}/${x}.${aext}" ]]; then
	  ffmpeg -y -i "${moviedir}/${x}.${vext}" -i "${moviedir}/${x}.${aext}" -c:v copy -c:a copy "${moviedir}/${x}.${voutext}" &&  rm -f "${moviedir}/${x}".{$aext,$vext,pid}
	  fi
  done <<< "${audiopids}"
  if ! pgrep -f "sudo nohup motion"; then sudo pkill -9 -f "arecord -f cd -r 22050 -D plughw:2,0"; fi
}

case ${action,,} in
stop ) stopaudio ;;
start ) startaudio ;;
esac

[ceewanna]

Based on your approach, I have tried recording parallel video files (with audio) with some loggings and found that the parallel videos are actually longer than those generated by motion. It is driven by event_gap. The on_event_end will kick in only when the event_gap has past which is way after the actual event ends. For instance, the default event_gap is 60s, then all the videos will have at least that event_gap built-in at the end. I am not so sure you have any problem reconcile your audio recorded with the intrinsic video generated by motion.

[ReedRYoung]

Nice work meirlazar, but when I tried to use motion to open a .WAV file in aplay "on_motion_detected", one of several problems I encountered was that PulseAudio didn't love references to specific sound devices, such as the "plughw:2,0" in your motionaudio.sh script. So I'm wondering, is there a more hands-off way to pipe both the audio and video more directly to a software package that's intended for A/V recording, and let that software handle all or most of the details in your "startaudio" and "stopaudio" functions, except for filepath?

[meirlazar]

I've added this hack to a github project - feel free to download, test, and submit any feedback/updates. Thanks everyone.
I've made some changes to it, so hopefully that resolves most of the issues or complaints I've heard so far...
https://github.com/meirlazar/motion_with_audio_arecord

[jlg89]

I'll put in a note of agreement that it would be super useful to have something like an "audio_passthru" setting (takes the audio from the "netcam_url" and passes it through to the output stream), or an "audio_url" setting (takes a separate audio stream and combines it with the output audio), because motion is such a great mediator between "dumb" RTSP cameras and various home automation systems. After all, "because it's difficult" is a reasonable argument for excluding a feature, but "because something someone might do with it might be illegal someplace" really isn't. Apple, for example, doesn't seem to be concerned that HSV drops a gajillion full A/V surveillance recordings into iCloud every day.

[ceewanna]

Motion is a great tool in terms of motion detection, especially the feature in which it uses low resolution for detection and passing through high resolution to video recording. This is very efficient compared to another OpenCV app that I have also setup to compare performance. Motion is very CPU friendly while dvr-scan based on OpenCV requires substantial much more CPU or even needs GPU. The only missing piece here is voice/sound recording.

You can always argue about legal issue. In my view, if you live in the context of no-sound-recording allowed, you just get all your mp4 through ffmpeg to strip off audio. That is it. Technically, it is just an option where you leave it to user's judgment to switch it on or off.

I live in a farm and I have Motion recording so many birds hanging around in my farm. The video recording is perfect but it's just too bad and really bad not being able to listen to their voice. One so called legal issue just kills off those lovely birds singing.

[pfzim]

Thanks for a great product!
But there is a lack of sound recording.
Please add optional pass-through audio recording from RTSP stream.