Closed hchaudhary1 closed 6 years ago
Yes I know, this is to decrease latency. But using 320 bytes creates an awful lot of overhead per packet and leads to throughput issues on the ESP32, so by trial and error I arrived at the current values.
got it. it looks like, you are transmitting as many samples as possible. I suppose it makes sense to let AVS buffer the data rather than an extremely constrained device like an ESP32
Per the AVS documentation, it is recommend to upload 10ms aka 320 bytes per chunk while uploading audio.
reviewing the code, it seems that this is not observed: https://github.com/MrBuddyCasino/ESP32_Alexa/blob/master/components/alexa/alexa_speech_recognizer.c#L115