Tacotron2 based multi-speaker text to speech

• Deep voice 2
• NVIDIA tacotron 2 implementation

Three places adding speaker embedding layers

• text encoder GRU initial state
• add speaker code after text encoder RNN

data:dictionaries for different language
dataset:where put training data and test data

python -m multiproc.py train.py -o outdir15/ -l logdir -c ./outdir5/tacotron2_statedict.pt --warm_start

location: 115 outdir14 data:

• mix indirect and direct speech with one speaker code + 39 female speaker from cusent
• storytelling sentences share the same speaker id

location: 114 outdir15 data:

• mix indirect and direct speech with one speaker code + 39 female speaker from cusent
• indirect speech: spk id 69 direct speech: spk id 70

• apex segmentation error: conda install -c psi4 gcc-5

• tweak input format : replace with pandas dataframe instead of filelists