djsamseng / browserlearner Goto Github PK

$ pacmd load-module module-loopback latency_msec=5pav
$ pavucontrol
# Input Devices
# Show All Input Devices
# Click the green check mark next to "Monitor of Built-in Audio Analog Stereo"

import pocketsphinx
for phrase in pocketsphinx.LiveSpeach():
    print(phrase)

Now play some sound (ex: ./data/recordings/a.wav) and it should print out what it heard

$ python3 record.py hello # saves to ./data/recordings/hello.wav

If you run into "OSError: [Errno -9998] Invalid number of channels" and it should work, try a few more times and eventually it works. Additionally you can run $ sudo killall pulseaudio

If you run into "OSError: [Errno -9981] Input overflowed" this may be because the microphone volume is too high. Turning down the microphone gain may solve this issue

1. Download and install sphinxbase and pocketsphinx per instructions
2. sudo apt-get install python-dev to fix "Could not link test program to Python. Maybe the main Python library has been installed in some non-standard library path. If so, pass it to configure, via the LDFLAGS environment variable."

$ cd pycli
$ python -m venv env
$ source ./env/bin/activate
$ pip install -r requirements.txt

$ cd ./server && yarn install

$ cd ./server && yarn start

$ cd ./client && yarn start
# open localhost:3000 in Chrome

$ cd pycli
$ source ./env/bin/activate
$ python3 main.py --letters
$ python3 main.py --youtube

$ conda create --name torch python=3.8
$ conda activate torch
$ python wav2vec_test.py