Microphone problem when attempting real-time SST conversion with the Riva C++ client

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

Hello @srella @rvinobha Linux Riva GPU - Hardware , I have this issue. I’m following the Riva client riva_streaming_asr_client, which runs successfully for examples that work with the direct [name].wav file.

Now, I’m trying to execute the “riva_streaming_asr_client” version to use the microphone and get live results as I speak, but I’m encountering this error:

COMMAND
./bazel-bin/riva/clients/asr/riva_streaming_asr_client --audio_device=hw:1,0

ANSWER/ERROR
I0117 19:51:01.987900 14380 grpc.h:94] Using Insecure Server Credentials
latency 100000
ALSA lib pcm.c:8664:(snd_pcm_set_params) Rate doesn’t match (requested 16000Hz, get 44100Hz)
snd_pcm_set_params error: Invalid argument
Error opening capture device hw:1,0

Could you please help me?

NOTE: I switched from Python to C++ due to a similar error , and in another post, they suggested using C++. Is there a solution for that error in Python now?

Hardware - GPU: NVIDIA GeForce RTX 3060
Hardware - CPU: 12th Gen Intel(R) Core™ i5-12400
Operating System: Ubuntu 22.04.3 LTS
Riva Version: 2.14.0

2

Hi @jbps.work

Thanks for your interest in Riva

I will check with the Engineering and get back

Thanks

3

Hi @jbps.work

Thanks for waiting

could you please try using --audio_device=plughw:1,0 that will ensure rate conversion if your recording device doesn’t support 16kHz

Thanks