README.md

GCloudSpeech

Transcribe voice data from stdin to text using Google Cloud Speech-to-Text.

Introduction

GCloudSpeech is a small application that allows you to parse voice data quick and easy using the command line.

This project can be used for Rhasspy, an offline, multilingual voice assistant toolkit. It's not an offline voice assistant anymore when you're using Google STT but it's worth a shot if you need good automatic speech recognition on a low-end device.

Installation

1. Clone this repository

   $ git clone https://github.com/NullEnt1ty/GCloudSpeech
   

2. Setup the virtual environment

   $ ./setup-venv.sh
   

Usage

You can transcribe voice data by piping it to run.sh. The transcription will be printed on standard output.

For example:

$ cat podcast.wav | ./run.sh --language en-US

Attention: Currently the only accepted format for voice data are uncompressed 16-bit signed little-endian samples (Linear PCM) with a sample rate of 16 kHz. This might be configurable in the future.

Integration into Rhasspy

Use the following configuration for your profile to integrate GCloudSpeech into Rhasspy:

"speech_to_text": {
    "command": {
        "program": "<path to run.sh>",
        "arguments": ["--language", "<your language code>"]
    },
    "system": "command"
},