Can RIVA offline models be used for language recognition with a microphone?

Hardware - AGX Orin
Hardware - CPU
Operating System:Ubuntu 20.04
Riva Version :2.8.1

Can RIVA offline models be used for language recognition with a microphone?
Are there any sample programs available for C++?

HI @u9713112

Thanks for your interest in Riva

Yes, of course, we can use RIVA offline models for language recognition, for ASR i.e speech to text,
For more details on ASR, please refer the below link

Once the Riva Model is deployed in your server, it can be access via C++ Client, Please find the repo below

Hope this helps


HI rvinobha

I saw the example of riva_asr_client from this website( GitHub - nvidia-riva/cpp-clients: Sample C++ command-line Riva clients.), but it can only perform speech recognition from a wav file. How can I modify it to use a microphone?


Hi @rvinobha
I have exact same question, that’s why pinging here, But required in “PYTHON”

Every thing is available, Deployed, Riva ready to use.

But How can I use streaming service from a websocket to transcribe audio buffer in real time. (Not In device microphone, Not pre recorded file).

Really looking for any help. Thank you.

HI @u9713112 and @ahmedshahzadaaa

Thanks for your reply,

I will further check with the team on how to live microphone instead of recorded files


1 Like

I’m waiting for your message~ :)
If possible, could you provide me with a C++ example?


@rvinobha Same waiting for you for python…

Hi @u9713112 and @ahmedshahzadaaa

Sincere Apologies for the delay

For live microphone use case using riva_streaming_asr_client with a streaming model would be the apt option, you can use --audio_device which will allow you to select the audio device to use.


Does the riva_asr_client example support the use of a microphone in non-streaming mode?
I want to implement riva asr functionality using a microphone without an internet connection.