Losing voice when two people speaking at the same time

Hi, thanks for the wonderful lib! When trying it, I found a problem as follows:

Let two persons (A and B) join an audio-only room. Let them use the speaker instead of earphones to play the audio. Then let both of them speak (e.g. A says 123456789 and B says 987654321). Then each person cannot hear the other guy’s voice (e.g. B will hear “1–novoice–3--novoice–7--novoice” instead of “1-2-3-4-5-6-7-8-9”, and similar for A).

To know whether it is a problem of networking, I tried another popular app, and also tried to use phone calls, and neither have such problems. So it may not be a networking problem.

I am trying to persuade my boss into using it, but this issue seems to be a big barrier…

Thanks very much for your reply!

P.S. There seem some posts with similar titles but the content are different.

Are you doing this in the same room, if so try separate the speakers and see how it goes?
Sound to me like echo cancelation and echo surpression thing, we do not control this as this is internal to the browser and webrtc, so if this is the case the issue is for the web browser you use.

Not same room, and the two testers are far away enough that they cannot hear each other even if shouting…
I also suspect the echo surpression, so it is sad to hear that you cannot modify it :frowning: I originally thought that is some algorithm you wrote…
I use the Android App for both testers, which run on a not-too-old Android system.

So is there anything that I can do to improve it? Thanks very much!

By the way, I find that it takes data of 15~20KB/s, while another similar app only takes ~8KB/s… So is there any way to make it better? Thanks!

For mobile webrtc part we use is https://github.com/react-native-webrtc/react-native-webrtc you can check it.
It uses opus as audio codec, maybe the best in terms of bandwidth, quality, and error correction abd default one when using webrtc. The 8 KB you see are those using opus?

Thanks for the information! I tried the WeChat audio call, which is a popular app and is closed source so I have no idea…