资源算法FFTNET_gaussian

FFTNET_gaussian

2020-01-17 | |  46 |   0 |   0

FFTNet

A PyTorch implementation of the FFTNet: a Real-Time Speaker-Dependent Neural Vocoder

Quick Start:

  1. Install requirements:

pip install -r requirements.txt
  1. Download dataset:

wget http://festvox.org/cmu_arctic/cmu_arctic/packed/cmu_us_slt_arctic-0.95-release.tar.bz2
tar xf cmu_us_slt_arctic-0.95-release.tar.bz2
  1. Extract features:

python preprocess.py
  1. Training with default hyperparams:

python train.py
  1. Synthesize from model:

    python generate.py --checkpoint=/path/to/model --lc_file=/path/to/local_conditon

TODO:

  • Test and modify conditonal sampling.

  • Fast generation. (Without fast generation, it generates about 230 samples per second with a free TITAN Xp.)

  • Post-synthesis denoising.

Notes:

  • This is not offical implementation, some details are different from the paper.

  • Now the generated speech from this repo is not so good as wavenet vocoder.

  • Work in progress.


上一篇:pytorch_FFTNet

下一篇:fftnet_subband

用户评价
全部评价

热门资源

  • TensorFlow-Course

    This repository aims to provide simple and read...

  • seetafaceJNI

    项目介绍 基于中科院seetaface2进行封装的JAVA...

  • mxnet_VanillaCNN

    This is a mxnet implementation of the Vanilla C...

  • DuReader_QANet_BiDAF

    Machine Reading Comprehension on DuReader Usin...

  • Klukshu-Sockeye-...

    KLUKSHU SOCKEYE PROJECTS 2016 This repositor...