Anyone have any recommendations on serving an rtsp audio stream from a raspberry pi?

[have-no-clue-what-im-doing]

I've been doing a bunch of research but I am not finding in solid documentation on this. There a specific library or package anyone suggests?

[tphakala]

Following uses ffmpeg to capture audio from usb sound card and it uses MediaMTX (previously known as rtsp-simple-server) to publish stream for network clients

sudo apt install ffmpeg

Download and install MediaMTX

wget https://github.com/bluenviron/mediamtx/releases/latest/download/mediamtx_linux_amd64.tar.gz
tar xzvf mediamtx_*_linux_arm64v8.tar.gz
sudo mv mediamtx /usr/local/bin/

Edit configuration file mediamtx.yml

Add the following content to the end of configuration file

paths:
  stream:
    runOnInit: ffmpeg -f alsa -i default -acodec libmp3lame -f rtsp -rtsp_transport tcp rtsp://localhost:8554/stream
    runOnInitRestart: yes

Start MediaMTX with the configuration file

mediamtx mediamtx.yml

Stream is now available at rtsp://your-pi-ip-address:8554/stream.

If you get errors about not being able to open device try these for ffmpeg -i default setting instead

-i sysdefault
-i pulse

[have-no-clue-what-im-doing]

Thanks @tphakala !

I did get this working with a few minor changes:

First, https://github.com/bluenviron/mediamtx/releases/latest/download/mediamtx_linux_amd64.tar.gz was giving me a 404 and it's amd64, so I went ahead and downloaded the arm version: https://github.com/bluenviron/mediamtx/releases/download/v1.8.3/mediamtx_v1.8.3_linux_arm64v8.tar.gz

And for the mediamtx.yml I kept getting errors about the "no device" no matter the setting, so I set the -i switch to hw:1,0 after running arecord -l to get a list of hardware devices.

I also had to set my channel to 1 since it kept trying to use 2 and erroring out. Here is what I added to my mediamtx.yml:

paths:
 stream:
  runOnInit: ffmpeg -f alsa -channels 1 -i hw:1,0 -acodec libmp3lame -f rtsp -rtsp_transport tcp rtsp://localhost:8554/stream
  runOnInitRestart: yes

After all this I finally got it working. Just putting this here for reference to anyone else having trouble setting this up.

[isZumpo]

Hi, I see that @tphakala already responded. I am using a similar setup in my Kubernetes cluster, with containers instead. Here are the relevant snippets of my pod specifications that could be translated into docker easily if you want to use it:

RTSP client running on Raspberry Pi with mic

 - image: lscr.io/linuxserver/ffmpeg:6.1.1
    name: ffmpeg
    args:
    - -f
    - alsa
    - -i
    - plughw:0
    - -f
    - "rtsp"
    - rtsp://birdnet-rtsps-server.birdnet.svc.cluster.local:8554/mystream
    env:
    - name: ALSA_CARD
      value: '1'

RTSP server which client talks to

 - image: bluenviron/mediamtx:latest
    name: mediamtx
    env:
    - name: MTX_PROTOCOLS
      value: tcp
    ports:
      - name: rtsp
        containerPort: 8554
        protocol: TCP

Birdnet

  args:
  - realtime
  - --rtsp
  - rtsp://birdnet-rtsps-server.birdnet.svc.cluster.local:8554/mystream
  - --rtsptransport
  - tcp

Swap the birdnet-rtsps-server.birdnet.svc.cluster.local address for whatever ip you decide to run the rtsp server at.

Bit of midsummer celebrations here in Sweden today. So I did unfortunately not have time to translate these into docker commands for you. But let me know if you need assistance with it 😊

[matthew73210]

I use the setup but went the vlc route and a cron to restart the server as sometimes it hangs

[alexbelgium]

You can have a better analysis sound quality by using PCM instead of mp3 (mind the conversion from 16 bits little endian to big endian needed for rtsp support) : ./mediamtx & ffmpeg -nostdin -f alsa -acodec pcm_s16le -ac 2 -ar 48000 -i hw:1,0 -f rtsp -acodec pcm_s16be rtsp://localhost:8554/birdmic -rtsp_transport tcp (I am using also rtsp conversion using this guide : mcguirepr89/BirdNET-Pi#1006 (comment))

[thor0215]

I posted this How To, Using VLC as a RTSP Server

[have-no-clue-what-im-doing]

Super helpful. Thank you!

[thor0215]

Feel free to add it to your instructions if you'd like.

[xconverge]

Personally I use go2rtc https://github.com/AlexxIT/go2rtc on a pi zero w

My config looks like this:

streams:
  mic:
    - exec:ffmpeg -hide_banner -f alsa -channels 1 -sample_rate 48000 -i hw:0,0 -c:a libopus -rtsp_transport tcp -f rtsp {output}