Closed AlexPeng19 closed 5 years ago
Have you seen this?
2017-06-28: The sample client program can now accept audio from stdin. This can be used to test the server with a live microphone, e.g.: arecord -f S16_LE -r 16000 | python kaldigstserver/client.py -r 32000 -. Thanks to @wkuna!
It should be resolved? Reopen if there is still a problem.
I am trying to use with live microphone using following command
_arecord -f S16LE -r 16000 | python kaldigstserver/client.py -r 32000
but getting following error..
_usage: client.py [-h] [-u URI] [-r RATE] [--save-adaptation-state SAVE_ADAPTATION_STATE] [--send-adaptation-state SEND_ADAPTATION_STATE] [--content-type CONTENTTYPE] audiofile client.py: error: too few arguments Recording WAVE 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
You are missing "-" at the end of the command.
Thanks Its working perfectly.. If I want to use my acoustic then which model I have to train? like tri2b, tri2b_mmi.. if neural network then which NN like nnet1, nnet2, nnet3... If possible tell me example in kaldi/egs.. Thank you in advance..
The tool supports trigram models (with no SAT), nnet2 models and nnet3 models. I recommend to use chain TDNN models.
Here I have complete and working code for decoding from microphone, python 2 and 3 are supported. Hope this help someone
i noticed the demo http://bark.phon.ioc.ee/dictate/ could get audio from microphone, but current README only showed the recognization by sending audio file. which api could i use to get that.
by the way i am using https://kaljurand.github.io/dictate.js, i want to replace wss://bark.phon.ioc.ee with my service.
looking forward your response.