Nvidia RIVA inferernce ASR real time

AI & Data Science Deep Learning (Training & Inference) Riva
1

Please provide the following information when requesting support.

Hardware - GPU T4
Hardware - CPU
Operating System Ubuntu
Riva Version V1.10
TLT Version (if relevant)
How to reproduce the issue ? (This is for errors. Please share the command and the detailed log here)

I am trying to transcribe a real time audio stream using Nvidia Riva client by fetching a websocket stream from a twilio voice call. The audio chunks I receive from the twilio stream return “No Response” from the Nvidia Riva service hosted on a AWS EC2 instance.

When I use the same setup with Google Speech to Text client, I get the response text on same audio chunks, from the same websocket stream.

2

Hi @sdm.amansehgal ,

Thanks for your interest in Riva,

Apologies you were facing some issues,

Can you please share the complete log, It would be helpful for us to find the cause and work in resolving your issue

3

@rvinobha ,

Thanks for reaching out, I figured out the issue. It was due to the encoding from on the twilio stream which uses MULAW. I had to make changes to the config to fix it.