DeepSpeech

Custom/fine-tuning models

See here

DStest.py

This tool can be used to query a deepspeech-server instance using various filtering mechanisms. For de-noising, requires the xipn rnnoise library to be compiled and installed and this to be locally available (also included in this repo). For the future, will add filter to make the silence padding optional as well.

usage: DStest.py [-h] [-H HIGHPASS] [-L LOWPASS] -U URL [-D] [-N] 
                 [-O ORDER] [-v] [--targetdb TARGETDB]
                 wavfile

positional arguments:
  wavfile               wav file to test on.

optional arguments:
  -h, --help            show this help message and exit
  -H HIGHPASS, --highpass HIGHPASS
                        high pass frequency.
  -L LOWPASS, --lowpass LOWPASS
                        low pass frequency.
  -U URL, --url URL     Deepspeech Server URL.
  -D, --denoise         de-noise clip.
  -N, --normalize       normalize clip.
  -O ORDER, --order ORDER
                        order of filters. Optional. HLDN would be Highpass,
                        Lowpass, denoise, then normalize. If order is
                        specified, only items included are run. N = normalize
                        D = de-noise L = low pass filter H = high pass filter
  -v, --verbose         Be verbose
  --targetdb TARGETDB   Decibel target for normalize. (optional)

Typical usage:

$ DStest.py  -N -D -L 3000 -H 300 --order NHLD -U http://localhost:1880/stt $wavfile
Filtered: you might have to say that a different way

Other things

Splitting a large file into separate wavs (adjust .3 as needed): sox input.wav output.wav silence 1 0.1 1% 1 0.3 1% : newfile : restart

Converting a file to the default DS format: sox input.wav -c 1 -b 16 -r 16000 output.wav