Martin Zheng

slope for last activation is different from the rest ones

2

Hi, The last activation slope in generator is the default value 0.01, while the rest ones are 0.1. Is this just a accident or you have a design consideration ?...

avoid hard-coding for generator input mel dimensions

hifigan vs tfgan

1

Hi, Thanks for your great work. did you compare the audio sample quality between hifigan v1 and tfgan?

你好，我在mac上编译你的项目，提示如下错误： c++ '-DNODE_GYP_MODULE_NAME=binding' '-DUSING_UV_SHARED=1' '-DUSING_V8_SHARED=1' '-DV8_DEPRECATION_WARNINGS=1' '-DV8_DEPRECATION_WARNINGS' '-DV8_IMMINENT_DEPRECATION_WARNINGS' '-D_DARWIN_USE_64_BIT_INODE=1' '-D_LARGEFILE_SOURCE' '-D_FILE_OFFSET_BITS=64' '-DOPENSSL_THREADS' '-DBUILDING_NODE_EXTENSION' -I/Users/martinzheng/.node-gyp/12.1.0/include/node -I/Users/martinzheng/.node-gyp/12.1.0/src -I/Users/martinzheng/.node-gyp/12.1.0/deps/openssl/config -I/Users/martinzheng/.node-gyp/12.1.0/deps/openssl/openssl/include -I/Users/martinzheng/.node-gyp/12.1.0/deps/uv/include -I/Users/martinzheng/.node-gyp/12.1.0/deps/zlib -I/Users/martinzheng/.node-gyp/12.1.0/deps/v8/include -I../../nan -I../src/libsass/include -Os -gdwarf-2 -mmacosx-version-min=10.7 -arch x86_64 -Wall -Wendif-labels -W...

multiband-hifigan

10

Hi, Did you try the idea multiband hifigan?

Hi, This is a nice work! Did you test the inference speed? how many tokens/second the model can generate on your GPU? i want to see if this model can...

what dataset you use for vocals_mdx23c model？

7

Hi, What dataset you use for vocals_mdx23c model training ? I want to reproduce your training and based on which to make improvement.

ISTFT head use center padding but the same padding

8

Hi this is a great work. I noticed that the ISTFT head does not use the same padding that you implemented, but uses the center padding which is provided by...

support pytorch-lightning 2 or above.