Issue feature request streaming decoder fast ds intermediatedecode calls

[JRMeyer]

>>> rajneeshy568
[September 2, 2019, 3:29pm]

With reference to this
issue and the pull
request that closes
it. slash
I am wondering if there is a implementation in the python interface of
DeepSpeech. I am not very good with C++. I have been looking to find a
sample implementation of the streaming recognizer for Deep speech's
python interface. Can someone please point me in the right direction?

Thanks a lot! slash
Yours

[This is an archived TTS discussion thread from discourse.mozilla.org/t/issue-feature-request-streaming-decoder-fast-ds-intermediatedecode-calls]

[JRMeyer]

>>> reuben
[September 2, 2019, 3:37pm]

The API is the same, the only difference is that DS_IntermediateDecode
is now fast enough to be usable.

[Archived Post]

[JRMeyer]

>>> eggonlea
[September 3, 2019, 7:55pm]

The python function which calls DS_INtermediateDecode is
intermediateDecode()

[Archived Post]

[JRMeyer]

>>> Abhinav
[September 4, 2019, 5:02am]

Here is a sample code showing this:

from scipy.io import wavfile
fs, frames = wavfile.read('abc.wav')
stream_context = ds.setupStream()
start = 0
start_time = time.time()
while(start<frames.shape[0]):
ds.feedAudioContent(stream_context, np.frombuffer(frames[start:start+4000], np.int16))
start += 4000
print('Done '+str(start/16000)+':'+str(time.time()-start_time)+' seconds')
print('Recognized: %s' % ds.intermediateDecode(stream_context))
text = ds.finishStream(stream_context)
print('Recognized: %s' % text)

[Archived Post]