Tihu is an open source Persian text-to-speech engine. It's a cross-platform application and mostly is written in C++. Tihu uses Hazm for digesting Persian text and Tihu g2p-seq2seq for Grapheme-to-Phoneme conversion.
Please make sure you have installed gcc (6.0 or higher). To check your gcc version run gcc --version.
You can compile Tihu by following these steps:
git clone https://github.com/tihu-nlp/tihu.git
cd tihu
# install hazm
pip install hazm
#compile MBROLA
git clone https://github.com/numediart/MBROLA
cd MBROLA
make
cp mbrola ./build/
# espeak-data
curl -sLO http://sourceforge.net/projects/espeak/files/espeak/espeak-1.48/espeak-1.48.04-source.zip
unzip espeak-1.48.04-source.zip
cp -r espeak-1.48.04-source/espeak-data ./build/data/
rm espeak-1.48.04-source.zip
rm -rf espeak-1.48.04-source
# g2p-seq2seq-tihudict
curl -sLO https://github.com/tihu-nlp/tihudict/releases/download/v1.0/g2p-seq2seq-tihudict-model-1.0.tar.gz
tar zxvf g2p-seq2seq-tihudict-model-1.0.tar.gz
mkdir ./build/data/g2p-seq2seq-tihudict
cp -r g2p-seq2seq-tihudict/* ./build/data/g2p-seq2seq-tihudict
# g2p-seq2seq-model-6.2-cmudict
curl -sLO https://sourceforge.net/projects/cmusphinx/files/G2P%20Models/g2p-seq2seq-model-6.2-cmudict-nostress.tar.gz
tar zxf g2p-seq2seq-model-6.2-cmudict-nostress.tar.gz
mkdir ./build/data/g2p-seq2seq-cmudict
cp -r g2p-seq2seq-model-6.2-cmudict-nostress/* ./build/data/g2p-seq2seq-cmudict
# compiling code
make
make console
Tihu console
Tihu console is a gui application that allows you to work with Tihu library. Before compiling Tihu console please make sure you have installed qt framework. You can also find a pre-compiled version of Tihu console in release page for linux x64.
Note: Windows users can compile Tihu by their own risks.
Tihu is Persian name for Partridge.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent