pydeepspeech 1.1.6

Mozilla's DeepSpeech transcriber in a pip installable package.

pydeepspeech

• The simpliest way to use AI to generate transcriptions from a wav file.
• This project uses the Mozilla DeepSpeech engine built from the included demo:
  • https://github.com/mozilla/DeepSpeech-examples/tree/r0.9/vad_transcriber

Quick start

Install to current python environment

Console api:

$ pip install pydeepspeech
$ pydeepspeech --wav_file <WAVE_FILE> --aggressive 1 --out_file <TEXT_FILE>

-or-

$ pydeepspeech --wav_file <WAVE_FILE> --out_file <TEXT_FILE> --model_dir <MY_PBMM_AND_SCORER_FILES>

-or-

 $ pydeepspeech_installmodels --pbmm <PBMM_FILE_OR_URL> --scorer <SCORER_FILE_OR_URL>
 $ pydeepspeech --wav_file <WAVE_FILE> --out_file <TEXT_FILE>

Or in python

from pydeepspeech.transcribe import transcribe
transcribe(...)

Optional: Create a virtual python package

Download and install virtual env:

# Download
curl -X GET https://raw.githubusercontent.com/zackees/make_venv/main/make_venv.py -o make_env.py
python make_env.py  # Make the environment
source activate.sh  # Enter environment
pip install -e .

To get back into the environment execute source activate.sh (if windows, you must be using git-bash)

Testing

Testing and linting is very simple. Just run tox

$ tox