I'm using this algorithm to get non-silent speech, but it's not recognized by gcp.

[cumthxy]

I'm using this algorithm to get non-silent speech, but it's not recognized by gcp.

i have a 60 seconds voice. and i use get_speech_timestamps to get nosound voice .but can't speech to text by gcp.
and the raw 60seconds voice can speech to text by gcp.

it is sampling_rate is 8000.

[snakers4]

Hi,

Please send your audio.
Also what is GCP?

[cumthxy]

gcp is a goole speech to text service.
this is the raw wav. the vad algorithm get_speech_timestamps can split. after that can't speech to text by gcp.
raw_wav.zip

[cumthxy]

I found many examples of this. but can't found how to solve it.
this also a case . the raw wav can speech to text by gcp or azure service.
raw_wav1.zip

but after i use this:
wav = read_audio(input_file, sampling_rate=8000) speech_timestamps = get_speech_timestamps(wav, model, sampling_rate=8000) if len(speech_timestamps) > 0: save_audio(output_file, collect_chunks(speech_timestamps, wav), sampling_rate=8000)

the output_file wav can't speech to text by gcp or azure service..
this is the processed wav.
text_wav.zip