Free and offline foreign (non-English) speech recognition with Python, Google API vosk and SpeechRecognition / Pocketsphinx.

For speech recognition with timestamps see timestamps folder.

Read these instructions to know what library you need to setup and how to do it. They are also available here.

• Online speech recognition with Google API:
  • pip install SpeechRecognition
• Offline speech recognition with vosk:
  • pip install vosk
  • download vosk model, unzip it and specify path to the model in program
• Offline Speech Recognition with SpeechRecognition and Pocketsphinx:
  • pip install SpeechRecognition
  • python -m pip install --upgrade pip setuptools wheel
  • pip install --upgrade pocketsphinx
  • download foreign models for pocketsphinx, unzip and setup it

See overview jupyter notebook, which contains examples of all methods.

Open it with jupyter or see directly in a browser.

As any python script any of these tree scripts (.py files) can be run with the following command: python script_name.py parameter1, parameter2 ....

Every script has two parameters:

• first (required) - name of the .wav file to recognize
• second (optional) - name of the text file to write recognized text. If not specified, uses first_parameter.txt

Examples:

• python script_online_sr.py audio.wav (writes text in audio.txt)
• python script_online_sr.py audio.wav audio_outout.txt
• python script_vosk.py 'sounds\filename.wav' (recognize current_folder\sounds\filename.wav)
• python script_offline_sr.py 'D:\sounds\filename.wav'