A simple audio file transcriber that uses the Google Cloud Speech API for transcription.
Install this package with pip:
pip install git+https://github.com/agouil/podcast-transcriber.gitInstall SoX - Sound eXchange. If you're using Mac you can install through Homebrew:
brew install soxIf you're using Windows or Linux, download the binaries and installer from here.
You will need to have a Google API Key in order transcript audio. If you don't have one, then you need to sign up for the Google Cloud Speech API.
Set the Google API Key as an enviroment variable. You can simply run,
For UNIX:
export GOOGLE_API_KEY=<your-api-key>For WINDOWS:
SET GOOGLE_API_KEY=<your-api-key>Copy the example script below to a file named example.py.
import argparse
import podcast_transcriber
# parse the CLI arguments
parser = argparse.ArgumentParser(prog="python podcast_transcriber.py")
parser.add_argument("input_file", help="input audio file")
args = parser.parse_args()
podcast_transcriber.transcribe(args.input_file)Run the script with:
python example.py input_fileThe argument input_file is the input audio file URL. E.g. For a podcast, you can provide the MP3 file found in a podcast's RSS Feed.
Then the script downloads the file and converts it to smaller files of 40 seconds length each of raw audio bytes through SoX. This is the format the Google Speech API requires - 16-bit 16KHz Linear PCM.
For each file, the script gets the transcript from the Google Speech API. In the end, it concatenates the transcript chunks to a final output file inside the output directory.
For testing the script you can use this small audio file - https://archive.org/download/testmp3testfile/mpthreetest.mp3
Fork the repository, make necessary changes, run tests and submit a pull request.
Before running tests, install the necessary requirements with:
pip install -r requirements_tests.txtRun tests with nosetests command.
To submit any issues, raise an issue through the Issues Page