I hear your frustration. This is something that we are very aware is causing a lot of pain and we are working on it but in order to give you some perspective, here is what is happening a little bit behind the scene.
We have been able to isolate the problems. We know that not all devices are being affected. So the problem is not systematic. We have collected several users information to be able to troubleshoot and have been ab le to reproduce the problem internally.
So far, we have been able to identified the problem but not been able to root cause it. Here is our understanding:
- Alexa users do not have that problem. (it’s important and you will see why later)
- We are using an RTMP stream from the camera to a self hosted Wowza server.
- The Wowza server is transcoding the RTMP stream into RTSP for Alexa and into MPEG-DASH for Google Home.
- Google Home is known to have a very strict implementation of the video protocols.
- Using VLC, we are able to view the MPEG-DASH stream properly without dropping anything (video or audio)
- The number of parties involved to get that system up and running is 6. Yes 6 different companies and subsidiaries.
Knowing those points, we still have too many scenario that can be the source of the problem:
- Does Google have a bug in their rendering engine? We tried to troubleshoot this kind of issue in the past and the result was just that the video stream adheres strictly to the video standard. It takes also a very long time to be able to monopolize the right resources on Google side to get the troubleshooting required deep into the google nest hub. Reason being that we are using the standard display renderers.
- Does the Wyze camera has a bug? It’s possible that we are not sending the information in the RTMP stream properly. It is unlikely but possible. The library that we are using is a 3rd party library and not something that we have developed ourselves from what I understand… Moreover, it is working just fine with Alexa so the RTMP stream is most likely correct.
- Does Wowza has a bug? From all we have seen above, it sounds like it and something is happening in the transcoding. But again, are we getting the proper signal out of the RTMP streams for transcoding to MPEG-DASH? We don’t know for sure.
So what is our working theory: Wowza is not transcoding to Google’s standards. We had that in the past when we tried to implement HLS for Google Home but had to fall back on MPEG-DASH.
As a consequence, we have opened a high severity ticket with Wowza to try to get to a resolution. Problem is: we are at the mercy of Wowza Tech Support at this point. I have to spare you the details but that communications is not the easiest because of geography and contracts…
What’s next?
Short term: we are continuing to push Wowza to help us troubleshoot the problem and help us address the problem.
Mid & Long term: we are working on implementing WebRTC to replace the complete stack that we have currently in place. This would give us way more control over the entire stack and would also cut some of the systemic issues that we have with the current implementation while allowing new set of features (hello local connection and two way communication)!