Overview

Jasper is an opensource project designed to make the process of creating a self-standing "Alexis"-style assistant. Because the project hasn't seen any new commits since 2017, it's a pretty tough project to get working for experienced engineers and few have ever successfully completed the task. But the value of a standalone voice assistant that does not interact with 3rd parties is important. Thus the creation of this DIY (do it yourself) guide.

This guide (with special thanks to Steve Keukes who provided the original draft) steps through the process of building a fully-functional Jasper installation on a Raspberry Pi. The guide should work if the user can be trusted to type commands as written with no additional creativity added. Every repository used by this guide has been safely forked and linked within GitHub. All files used have been stored in an accompanying SourceForge project owned by the author of this guide (Adam Peacock) at https://sourceforge.net/projects/jasperclient/files/.

Work is underway to convert Jasper to Python 3 - the progress of which can be monitored HERE

Package details are listed HERE

Hardware Used in this Tutorial (tutorial -reportedly- works on a Model 4 as well)

• Raspberry Pi 3 B+ (https://www.amazon.com/gp/product/B07BDR5PDW)
• CanaKit Raspberry Pi 3 B+ (B Plus) with Premium Clear Case and 2.5A Power Supply (https://www.amazon.com/gp/product/B00MARDJZ4)
• Edimax EW-7811Un 150Mbps 11n Wi-Fi USB Adapter (https://www.amazon.com/gp/product/B003MTTJOY)
• Samsung 32GB 95MB/s (U1) microSDHC EVO Select Memory Card (https://www.amazon.com/gp/product/B06XWN9Q99)
• Conference Microphone, OmniDirectional Dual Capsules, USB Mic (https://www.amazon.com/gp/product/B07LC3FT5B)
• Speaker system (with 3.5mm jack) for playback

Jasper on Raspian Buster Software Installation Guide

This install process was tested on a Raspberry Pi Model 3 B+ and Model 4 using the Raspian Buster Lite ISO (https://sourceforge.net/projects/jasperclient/files/2020-02-13-raspbian-buster-lite.zip). The file can be downloaded from the public repo but to emnsure the file versions do not change, the IMG file was moved to SourcerForge to ensure the version of the initial OS is known/tested. This tutorial used a USB microphone and speakers that plug into the audio jack and uses PocketSphinx for the speech-to-text (STT) and Festival for text-to-speech (TTS). This method is fairly secure in that there is no traffic sent to 3rd parties.

Feeling Lazy?

A fully-functional Jasper installation was burned to an IMG file that works on any RPI 3B+ (possibly other models but they haven't been tested). The file can downloaded directly at https://sourceforge.net/projects/jasperclient/files/jasper-rpi.zip/download and written to an SD-card using any image burning software (i.e. Etcher). The username and password were not changed from the Raspbian defaults. (Username: pi Password: raspberry)

Installation Guide

Boot the Raspian image on the RPi device and sign in. Use raspi-config to enable the network adapter or wireless as needed and enable SSH to make thigns easier. Plug in the microphone and speakers then restart the Pi.

ssh pi@<ip-address>

Update and install some utilities

sudo apt-get -y update
sudo apt-get -y upgrade
sudo apt-get install -y vim git-core python-dev bison libasound2-dev libportaudio-dev python-pyaudio dos2unix
sudo apt-get install -y python-setuptools
sudo python /usr/lib/python2.7/dist-packages/easy_install.py pip

Create/edit ALSA configuration file:

sudo touch /lib/modprobe.d/jasper.conf &&  sudo chown pi /lib/modprobe.d/jasper.conf && sudo cat > /lib/modprobe.d/jasper.conf<<EOF
#Loads USB audio before the internal soundcard
options snd_usb_audio index=0
options snd_bcm2835
#snd_bcm2835 = built-in audio jack
#snd_usb_audio = USB-connected microphone
#Makes sure the sound cards are listed as desired in ALSA
#Verify order (and what ALSA sees) by entering "alsamixer" at the command line
#If a device is not listed, reboot the device and try again.
#options snd slots=<capture_device>,<playback_device>
options snd slots=snd_usb_audio,snd_bcm2835
EOF
sudo reboot

Test the microphone and speakers:

Note: This tutorial mentions alsamixer as a handy tool to test playback and microphone gain/volume. However, if certain aspects of alsamixer don't work quite right (i.e. F4 Capture), feel free to troubleshoot but don't add software binaries not mentioned in this guide unless you know what you're doing. * Remember this tool is a matter convenience - not a requirement.

• To start alsamixer: $ alsamixer
• Press F6 to select microphone
• Press F4 to capture
• Press up arrow to increase microphone gain
• To record: $ arecord test.wav, speak into the microphone, end recording: Ctrl + C
• To playback: $ aplay -D hw:1,0 test.wav

Edit & source profile scripts:

touch .bash_profile && cat>>.bash_profile<<EOF
export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/usr/local/lib:/usr/local/lib/fst"
EOF
source .bash_profile
touch .bashrc && cat>>.bashrc<<EOF
LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/usr/local/lib:/usr/local/lib/fst"
export LD_LIBRARY_PATH
PATH=$PATH:/usr/local/lib:/usr/local/lib/fst
export PATH
EOF
source .bashrc

The authoritative repo for jasper is at https://github.com/jasperproject/jasper-client.git but it hasn't been updated since 2017. This tutorial uses a forked repo:

git clone https://github.com/aplawson/jasper-client.git jasper
wget https://raw.githubusercontent.com/aplawson/jasperbuster/master/jasper.patch
cd jasper && patch -p1 < ../jasper.patch
cd

Install needed Python libraries

sudo pip install --upgrade setuptools
sudo pip install -r jasper/client/requirements.txt        #### this takes a while

Create the Jasper profile

cd ~/jasper/client
python populate.py
cd

Install PocketSphinx and some necessary utilities

sudo apt-get install -y pocketsphinx python-pocketsphinx pocketsphinx-en-us
sudo apt-get install -y subversion autoconf libtool automake gfortran g++

Install CMUSphinx

wget --no-check-certificate https://master.dl.sourceforge.net/project/jasperclient/cmuclmtk.zip
unzip cmuclmtk.zip && chmod 777 -R cmuclmtk && cd cmuclmtk/ && find . -type f -exec dos2unix -k -s -o {} ';'
./autogen.sh
make
sudo make install
cd

Download and untar Phonetisaurus, m2m-aligner and MITLM and OpenFST (needed for the Pocketsphinx STT engine)

wget --no-check-certificate https://master.dl.sourceforge.net/project/jasperclient/openfst-1.3.4.tar.gz
wget --no-check-certificate https://master.dl.sourceforge.net/project/jasperclient/mitlm_0.4.1.tar.gz
wget --no-check-certificate https://master.dl.sourceforge.net/project/jasperclient/m2m-aligner-1.2.tar.gz
wget --no-check-certificate https://master.dl.sourceforge.net/project/jasperclient/is2013-conversion.tgz
tar -xvf m2m-aligner-1.2.tar.gz
tar -xvf openfst-1.3.4.tar.gz
tar -xvf is2013-conversion.tgz
tar -xvf mitlm_0.4.1.tar.gz

Patch and build OpenFST.

wget https://raw.githubusercontent.com/aplawson/jasperbuster/master/openfst.patch
cd openfst-1.3.4 && patch -p 1 < ../openfst.patch
sudo ./configure --enable-compact-fsts --enable-const-fsts --enable-far --enable-lookahead-fsts --enable-pdt
  # Next step takes very long time. There might be some gcc warnings -- just ignore
sudo make install
cd

Build m2m-aligner

cd m2m-aligner-1.2/
sudo make
cd

Build mitlm

cd mitlm-0.4.1/
sudo ./configure
sudo make install
cd

Build Phonetisaurus

cd is2013-conversion/phonetisaurus/src
sudo make
cd

Move some files around

sudo cp ~/m2m-aligner-1.2/m2m-aligner /usr/local/bin/m2m-aligner
sudo cp ~/is2013-conversion/bin/phonetisaurus-g2p /usr/local/bin/phonetisaurus-g2p

Download and build the Phonetisaurus FST model

wget --no-check-certificate https://master.dl.sourceforge.net/project/jasperclient/g014b2b.tgz
tar -xvf g014b2b.tgz && mv ~/g014b2b ~/phonetisaurus && cd phonetisaurus
./compile-fst.sh
cd

Install Festival for TTS

sudo apt-get install -y festival festvox-don

Final edits the Jasper profile.yml

The use of Facebook and Gmail credentials are optional if email access or Facebook integration is desired

cat > .jasper/profile.yml<<EOF
carrier: carriername
first_name: first-name
gmail_address: emailaddress
gmail_password: emailpassword
last_name: last-name
phone_number: '1234567890'
prefers_email: false
stt_engine: sphinx
timezone: America/Los_Angeles
pocketsphinx:
    hmm_dir: '/usr/share/pocketsphinx/model/en-us/en-us'
    fst_model: '../phonetisaurus/g014b2b.fst'
tts_engine: festival-tts
EOF

Reboot

sudo reboot

Start Jasper

cd jasper
./jasper.py

PACKAGE DETAILS

Temporarily required for install/compiling process:

vim To edit files manually if needed

git-core To clone repositories

dos2unix To remove non-UNIX characters from some files that break functionality. Not needed after script is cleaned up

g++ To compile packages from source

python-setuptools/pip To install python 2.7 modules

automake To assist with make/compile process(es)

autoconf To assist with make/compile process(es)

Required for Jasper to function (post-install):

python-dev To install python-development tools used by Jasper.

WARNING: The Jasper project is written in Python 2 which is officially EOL (end-of-life) effective JAN-1-2020. The viability of this guide is dependent on the eventual migration from Python 2 to Python 3. Work is underway to convert to Python 3 - the progress of which can be monitored in the jasper-client-python3 branch.