Front end of audio cut off at start of call

[LynxEric]

Hello everyone, I'm dealing with an interesting issue relating to audio streaming. I'm using this library simply to transmit audio from my application to audio endpoints, much like how the PlaySounds example handles sending sound data via a stream. If I call a speaker directly, the front end of audio gets cut off. Its inconsistent, but very often the first half second or so of audio seems to get dropped on the floor during the call. Capturing SIP, SDP, and RTP with Wireshark reveals that the entirety of the sound file is transmitted, and the call flow appears normal (INVITE -> 200 OK <- ACK -> RTP starts) so I don't think the RTP stream starts up too quickly i.e the speaker doesn't have enough time to "wake up". Oddly, when registering the speaker to and placing the call thru a PBX like 3CX this issue does not occur, which tells me that 3CX is doing some magic I'm unaware of.
As a workaround I'm adding a half second of silence to the beginning of calls, with room for configuration if need be, but that feels like a bodge and not a real solution. I'm hoping you sorcerers can shed some light on what might be going on, or even just confirm that adding a buffer of silence is standard or acceptable practice.

[sipsorcery]

Are one or both ends of the call behind a NAT?

[LynxEric]

Yes! Both my SIP service and speaker are behind a NAT

[sipsorcery]

If they are behind different NATs where did you do the wireshark capture? Was it on the network that's experiencing the audio delay?

NAT is always the prime culprit when it comes to certain types of audio issues.

Alternatively you could try with bout UA's on the same network and see if the behavior is still present or not.