My first experiment with machine learning. I show you how to get and setup Mozilla Text To Speech neural network. How to learn it, and how to use.
I try to setup mozilla/tts (TTS – text to speech) to my MacBook Pro 2017. And I need to say, learning TTS can take few days… And if you have same laptop as I, you need more powerfull computer.
So. At first I make new VirtualBox instance of Debian 9 or 10 and x64. I think no need to explain how to do it. All you need it’s standard utilities and ssh server to install while you install debian Go to next step 😉
Let’s clone from git.
We need to correct environment var PATH. Execute
cat .profile > .bash_profile after that our ~./bash_profile must be same as lower block.
alias python='python3' alias python2='python2' # if running bash if [ -n "$BASH_VERSION" ]; then # include .bashrc if it exists if [ -f "$HOME/.bashrc" ]; then . "$HOME/.bashrc" fi fi # set PATH so it includes user's private bin if it exists if [ -d "$HOME/bin" ] ; then PATH="$HOME/bin:$PATH" fi # set PATH so it includes user's private bin if it exists if [ -d "$HOME/.local/bin" ] ; then PATH="$HOME/.local/bin:$PATH" fi
Let’s make some additional setup
cd # Just to go to our home directory su # To install GIT apt-get update apt-get upgrade apt-get install git # Say Y to all questions apt-get install pkg-config # for matplotlib apt-get install python3-pip # To use python3 package manager apt-get install python-dev python3-dev apt-get install build-essential libssl-dev libffi-dev libxml2-dev libxslt1-dev zlib1g-dev apt-get install libsndfile1 apt-get install python3-cairo apt-get install setuptools exit # Return to our non-privileged user pip3 install freetype-py pip3 install pypng cat requirements.txt pip3 install numpy==1.14.3 pip3 install 'torch>=0.4.1' pip3 install librosa==0.5.1 pip3 install Unidecode==0.4.20 pip3 install tensorboard pip3 install tensorboardX pip3 install Pillow pip3 install flask pip3 install tqdm pip3 install soundfile pip3 install matplotlib==2.0.2 # 2.2.2 with no errors wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh chmod 755 ./Miniconda3-latest-Linux-x86_64.sh ./Miniconda3-latest-Linux-x86_64.sh unlink ./Miniconda3-latest-Linux-x86_64.sh ~/miniconda3/bin/conda install -c ilastik scipy su -c 'chmod -R 755 ./TTS/' cd ./TTS su -c 'python3 ./setup.py develop' # we have error but i'ts ok pip3 install segments pip3 install attrs delete all rows from setup.py in dependency block except phonemizer get models or learn python3 ./synthesize.py
Now we need to setup default python version to 3 or higher. Read article in my blog > How to set Python’s default version
Next step is clone TTS from mozilla git repository.
git clone https://github.com/mozilla/TTS.git
Wait some time wile all files downloaded…
cd TTS # it's our main neural network directory su root -c 'python3 ./setup.py develop'
After we try to install by setup.py we have error.
* The following required packages can not be built: * freetype, png
We need to install it by hands.
pip3 install freetype-py pip3 install pypng
After we install all missed libraries try to setup another once
segments attrs joblist su -c 'apt-get install python3-matplotlib'
Version of matplotlib you can get from requirements.txt by command
Before we start install all things from requirements.txt or use setup.py like in README.md we need to install some additional things not listed in instructions.
Now we have python3 by default and we can start setup out environment for TTS.