Skip to content
Switch branches/tags
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Raspberry Pi app that continuously runs Google's Speech-to-Text Streaming API and relays it onwards via BTLE. Managed using MQTT.

This is a companion to the iOS app called Subtitles.

This was designed and tested to run on a Raspberry Pi Zero W using a ReSpeaker microphone array.


Getting Started

Setup Rasperry Pi Zero W


To setup the Raspberry Pi Zero W, first download the appropriate image and flash it to the Micro SD card:

  1. Download the Raspbian Stretch Lite image from here

  2. Download Etcher and install.

  3. Create an empty file with touch on the SD Card called ssh

e.g. touch /Volumes/boot/ssh

  1. Create an empty file with touch on the SD Card called wpa_supplicant.conf

e.g. /Volumes/boot/wpa_supplicant.conf

  1. Add your Wi-Fi details to wpa_supplicant.conf:
ctrl_interface=DIR=/var/run/wpa_supplicant GROUP=netdev

  1. Eject the SD card and insert in to the Pi Zero W. Boot up the machine and wait around 2 minutes.

For instructions on connecting to multiple Wi-Fi networks, see here.


To SSH in to the Pi Zero W

$ ssh-keygen -R raspberrypi.local

$ ssh pi@raspberrypi.local

Default password is raspberry, but it is advised you change this.


Installing Node.js for Raspberry Pi Zero W

The Raspberry Pi Zero uses a different ARM architecture (v6l) to the Raspberry Pi 2 & 3, therefore we need to install Node.js differently.

Remove any existing Node.js binaries and run the following commands:

$ wget

$ tar -xzf node-v7.7.2-linux-armv6l.tar.gz

$ sudo cp -R node-v7.7.2-linux-armv6l/* /usr/local/

$ echo 'PATH=$PATH:/usr/local/bin' >>  ~/.profile

Linux requirements

To ensure the Bluetooth works, please follow instructions here for Linux.

Checkout and build from source

Due to the architecture difference, we cannot rely on npm to install the right C++ binaries.

Therefore, to build the bobbycom repo, do the following:

$ git clone

$ cd bobbycom

$ cd network && npm install --build-from-source --save && cd ..

$ cd speech && pip install -r requirements.txt && cd ..

$ cd ui && pip install -r requirements.txt && cd ..

$ npm install -g forever

Add Google Credentials

To authenticate and use the Google Speech-to-Text API, you need to download your project's credentials file and copy it to the Pi e.g. the /home/pi/bobbycom directory.

You can use scp to copy it across. For example:

$ scp -r google-credentials.json pi@raspberrypi.local:/home/pi/bobbycom

Setup Microphone

To setup the ReSpeaker to work on the Raspberry Pi, create a file at ~/.asoundrc ($ touch ~/.asoundrc) and then add the following:

pcm.!default {
    type asym
    playback.pcm {
        type plug
        slave.pcm "hw:0,0"
    capture.pcm {
        type dsnoop
        ipc_key 5432
        slave {
            pcm "hw:1,0"
            buffer_size 96000


  • Network - relays ASR results over BTLE and periodically checks Internet connection

  • UI - handles UI control

  • Speech - Indefinite processing and streaming of audio and ASR


  • MQTT is automatically started on the system.

  • Forever is used to manage the processes.

Manually start:

$ sudo forever start /home/pi/bobbycom/forever.json &

Start at boot:

To start bobbycom at boot, copy the rc.local file to /etc/rc.local.

e.g. $ cp rc.local /etc/rc.local

Running Forever commands

If you start Forever via /etc/rc.local, it will be run as root, therefore you need to use sudo:

$ sudo forever <command>

Otherwise, you can use:

$ forever <command>


forever.json defines the processes to be run by Forever and lists all the log destinations.


  • Spinning GREEN LEDs = loading
  • Solid YELLOW LEDs = waiting for Internet connection
  • Solid BLUE LEDs = waiting for Bluetooth connection
  • Solid RED LEDs = error, requires reboot
  • Solid GREEN LEDs = ready, after they disappear, will be ready to recognize and relay speech.

Notes about boot sequence

Between powering on the Pi and the UI initiating the loading sequence (see below), the ReSpeaker will acknowledge audio with the GREEN|BLUE|GREEN pattern showing on the direction it hears audio from. This is because the firmware of the ReSpeaker is difficult to override and therefore isn't controlled until later in the boot sequence. Until you see the GREEN LED solid pattern, it will not recognize and relay speech.


Apache 2.0.


Raspberry Pi app that continuously runs Google's Speech-to-Text Streaming API and relays it onwards via BTLE.




No releases published


No packages published