Search code examples
node.jsffmpegsipvoip

Streaming RTP with ffmpeg and node.js to voip phone


I am trying to implement SIP in node.js. Here is the library i am working on

Upon receiving an invite request such as


Received INVITE
INVITE sip:[email protected]:5060 SIP/2.0
Via: SIP/2.0/UDP 192.168.1.39:5062;branch=z9hG4bK1534941205
From: "Nik" <sip:[email protected]>;tag=564148403
To: <sip:[email protected]>
Call-ID: [email protected]
CSeq: 2 INVITE
Contact: <sip:[email protected]:5062>
Authorization: Digest username="Nik", realm="NRegistrar", nonce="1234abcd", uri="sip:[email protected]:5060", response="7fba16dafe3d60c270b774bd5bba524c", algorithm=MD5
Content-Type: application/sdp
Allow: INVITE, INFO, PRACK, ACK, BYE, CANCEL, OPTIONS, NOTIFY, REGISTER, SUBSCRIBE, REFER, PUBLISH, UPDATE, MESSAGE
Max-Forwards: 70
User-Agent: Yealink SIP-T42G 29.71.0.120
Supported: replaces
Allow-Events: talk,hold,conference,refer,check-sync
Content-Length: 306

v=0
o=- 20083 20083 IN IP4 192.168.1.39
s=SDP data
c=IN IP4 192.168.1.39
t=0 0
m=audio 11782 RTP/AVP 0 8 18 9 101
a=rtpmap:0 PCMU/8000
a=rtpmap:8 PCMA/8000
a=rtpmap:18 G729/8000
a=fmtp:18 annexb=no
a=rtpmap:9 G722/8000
a=fmtp:101 0-15
a=rtpmap:101 telephone-event/8000
a=ptime:20
a=sendrecv

I can then parse the SDP into an object like this

 
{
    "session":{
        "version":"0",
        "origin":"- 20084 20084 IN IP4 192.168.1.39",
        "sessionName":"SDP data"
    },
    "media":[
        {
            "media":"audio",
            "port":11784,
            "protocol":"RTP/AVP",
            "format":"0",
            "attributes":[
                "rtpmap:0 PCMU/8000",
                "rtpmap:8 PCMA/8000",
                "rtpmap:18 G729/8000",
                "fmtp:18 annexb=no",
                "rtpmap:9 G722/8000",
                "fmtp:101 0-15",
                "rtpmap:101 telephone-event/8000",
                "ptime:20",
                "sendrecv"
            ]
        }
    ]
}

After sending the 100 and 180 responses with my library i attempt to start a RTP stream with ffmpeg

var port = SDPParser.parse(res.message.body).media[0].port
var s = new STREAMER('output.wav', '192.168.1.39', port)

with the following STREAMER class

class Streamer{
    constructor(inputFilePath, rtpAddress, rtpPort){
        this.inputFilePath = 'output.wav';
        this.rtpAddress = rtpAddress;
        this.rtpPort = rtpPort;
    }

    start(){
        return new Promise((resolve) => {
            const ffmpegCommand = `ffmpeg -re -i ${this.inputFilePath} -ar 8000 -f mulaw -f rtp rtp://${this.rtpAddress}:${this.rtpPort}`;
            const ffmpegProcess = spawn(ffmpegCommand, { shell: true });
    
            ffmpegProcess.stdout.on('data', (data) => {
                data = data.toString()
                //replace all instances of 127.0.0.1 with our local ip address
                data = data.replace(new RegExp('127.0.0.1', 'g'), '192.168.1.3');

                resolve(data.toString())
            });
    
            ffmpegProcess.stderr.on('data', (data) => {
              // Handle stderr data if required
              console.log(data.toString())
            });
    
            ffmpegProcess.on('close', (code) => {
              // Handle process close event if required
              console.log('close')
              console.log(code.toString())
            });
    
            ffmpegProcess.on('error', (error) => {
              // Handle process error event if required
              console.log(error.toString())
            });
        })
    }
     
}

the start() function resolves with the SDP that ffmpeg generates. I am starting to think that ffmpeg cant generate proper SDP for voip calls.

so when i create 200 response with the following sdp

v=0
o=- 0 0 IN IP4 192.168.1.3
s=Impact Moderato
c=IN IP4 192.168.1.39
t=0 0
a=tool:libavformat 58.29.100
m=audio 12123 RTP/AVP 97
b=AS:128
a=rtpmap:97 PCMU/8000/2

the other line never picks up. from my understanding the first invite from the caller will provide SDP that will tell me where to send the RTP stream too and the correct codecs and everything. I know that currently, my wav file is PCMU and i can listen to it with ffplay and the provided sdp. what is required to make the other line pickup specifically a Yealink t42g

my full attempt looks like this

Client.on('INVITE', (res) => {
    console.log("Received INVITE")
    var d = Client.Dialog(res).then(dialog => {
        dialog.send(res.CreateResponse(100))
        dialog.send(res.CreateResponse(180))
        var port = SDPParser.parse(res.message.body).media[0].port

        var s = new STREAMER('output.wav', '192.168.1.39', port)
        s.start().then(sdp => {
            console.log(sdp.split('SDP:')[1])
            var ok = res.CreateResponse(200)
            ok.body = sdp.split('SDP:')[1]
            dialog.send(ok)
        })

        dialog.on('BYE', (res) => {
            console.log("BYE")
            dialog.send(res.CreateResponse(200))
            dialog.kill()
        })
    })
})

I have provided a link to my library at the top of this message. My current problem is in the examples/Client folder.

I'm not sure what could be going wrong here. Maybe i'm not using the right format or codec for the VOIP phone i dont see whats wrong with the SDP. especially if i can listen to SDP generated by ffmpeg if i stream RTP back to the same computer i use ffplay on. Any help is greatly appreciated.

Update

As i test i decided to send the caller back SDP that was generated by a Yealink phone like itself. but with some modifications

v=0
o=- ${this.output_port} ${this.output_port} IN IP4 192.168.1.39
s=SDP data
c=IN IP4 192.168.1.39
t=0 0
m=audio ${this.output_port} RTP/AVP 0 8 18 9 101
a=rtpmap:0 PCMU/8000
a=rtpmap:8 PCMA/8000
a=rtpmap:18 G729/8000
a=fmtp:18 annexb=no
a=rtpmap:9 G722/8000
a=fmtp:101 0-15
a=rtpmap:1
01 telephone-event/8000
a=ptime:20
a=sendrecv

Finally, the phone that makes the call in the first place will fully answer but still no audio stream. I notice if I change the IP address or port to something wrong the other phone Will hear its own audio instead of just quiet. so this leads me to believe I am headed in the right direction. And maybe the problem lies in not sending the right audio format for what I'm describing.

Additionaly, Whenever using ffmpeg to stream my audio with rtp I notice that it sees the file format as this pcm_alaw, 8000 Hz, mono, s16, 64 kb/s My new SDP describes using both ulaw and alaw but I'm not sure which it is saying it prefers

v=0
o=- ${this.output_port} ${this.output_port} IN IP4 192.168.1.39
s=SDP data
c=IN IP4 192.168.1.39
t=0 0
m=audio ${this.output_port} RTP/AVP 0 101
a=rtpmap:0 PCMU/8000
a=fmtp:101 0-15
a=rtpmap:101 telephone-event/8000
a=ptime:0
a=sendrecv

I have been able to simply the SDP down to this. This will let the other phone actually pickup and not hear its own audio. it's just a completely dead air stream.


Solution

  • The problem turned out to be the ports in the SDP

    I originally thought that I should make up a port to use and set it in the SDP, Not that the initial SDP that was supplied to my client actually told me the port to stream on. I thought it was to listen on.

    After changing the ports to reflect this new discovery it worked. Although the audio is a bit choppy for Yealink phones. But for my Zoiper soft phones it's crystal clear. But that's for another day.

    v=0
    o=- 12420 12420 IN IP4 192.168.1.3
    s=SDP data
    c=IN IP4 192.168.1.3
    t=0 0
    m=audio 12420 RTP/AVP 0 101
    a=rtpmap:0 PCMU/8000
    a=fmtp:101 0-15
    a=rtpmap:101 telephone-event/8000
    a=ptime:0
    a=sendrecv