I've got some success with streaming audio from python's queue object. https://gist.github.com/datspike/3caabf60c04ffb726c319d70a59c4552
If you can dump audio from the mic into some queue.Queue() object and pass that to this class, and then use it like that:

with MockStreamFile(some_queue) as audio_file:
    self.audio_sample = self.speech_client.sample(
        stream=audio_file,
        encoding=speech.encoding.Encoding.LINEAR16,
        sample_rate_hertz=16000)

It will read data from python's queue as send it to the speech api without any problems.
I have a question, how can I get current streaming recognition state? I'm trying to get more or less correct timestamps from recognition results and I need some reference point aka at which time API has started to recognize the speech. Obviously I can get it from the

for alternative in alternatives:
    pass

...loop, but then i will lose all the time when audio did not have any speech in it.
Second path is to save time before this loop, but in my testing it's a bit not stable in terms of correct timestamps from time to time.

V1 Speech grpc streaming #932

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions