This repository contains the Neural Network (NN) based Speech Synthesis System
developed at the Centre for Speech Technology Research (CSTR), University of
Edinburgh.
To build the toolkit: see ./INSTALL
. These instructions are valid for UNIX
systems including various flavors of Linux;
To run the example system builds, see egs/README.txt
As a first demo, please follow the scripts in egs/slt_arctic
virtualenv venv -p python2
source venv/bin/activate
pip install -U numpy scipy theano lxml matplotlib bandmat
cd egs/slt_arctic/s1
./run_demo.sh
Listen to synthetic speech samples from our demo voice.
- Create a personal fork of the main Merlin repository in GitHub.
- Make your changes in a named branch different from
master
, e.g. you create a branchmy-new-feature
. - Generate a pull request through the Web interface of GitHub.
Post your questions, suggestions, and discussions to GitHub Issues.
If you publish work based on Merlin, please cite:
Zhizheng Wu, Oliver Watts, Simon King, "Merlin: An Open Source Neural Network Speech Synthesis System" in Proc. 9th ISCA Speech Synthesis Workshop (SSW9), September 2016, Sunnyvale, CA, USA