search

alokprasad / LPCTron

Tacotron2 + LPCNET for complete End-to-End TTS System
93 stars 43 forks source link

Any audio sample? #3

Open terryyizhong opened 5 years ago

terryyizhong commented 5 years ago

Hi thanks for the work. Wondering if it is possible to provide some audio sample synthesis from this repo. Or pretrained model.

alokprasad commented 5 years ago

You can try c++ project https://github.com/alokprasad/lpctron-tts-cpp it has model in model folder ( you to merge the files to generate the pb). Will upload samples later.

terryyizhong commented 5 years ago

You can try c++ project https://github.com/alokprasad/lpctron-tts-cpp it has model in model folder ( you to merge the files to generate the pb). Will upload samples later.

thanks for your quick reply. I've tried your c++ project. I got the pb model file first. after step2: mkdir build && cd build I do : cmake ..(you may miss this step in readme) make And error show: "cc: error: unrecognized command line option ‘-fvectorize’" when make I've tried 2 machine, both report this error. (I use CMAKE version 3.14.0) Did I do anything wrong?

alokprasad commented 5 years ago

can you modify Cmakelists.txt uncomment include(CMakeLists-static-linux.txt) and comment the android one.

terryyizhong commented 5 years ago

I just did what you said, the same error still occur. I don't very familiar with C++. full error log: [ 9%] Building C object lpcnet/CMakeFiles/lpcnet.dir/src/lpcnet.c.o cc: error: unrecognized command line option ‘-fvectorize’ lpcnet/CMakeFiles/lpcnet.dir/build.make:62: recipe for target 'lpcnet/CMakeFiles/lpcnet.dir/src/lpcnet.c.o' failed make[2]: [lpcnet/CMakeFiles/lpcnet.dir/src/lpcnet.c.o] Error 1 CMakeFiles/Makefile2:127: recipe for target 'lpcnet/CMakeFiles/lpcnet.dir/all' failed make[1]: [lpcnet/CMakeFiles/lpcnet.dir/all] Error 2 Makefile:83: recipe for target 'all' failed make: *** [all] Error 2

terryyizhong commented 5 years ago

May be I complied tensorflow wrong, I'll try again, looking forward your audio samples~

alokprasad commented 5 years ago

@terryyizhong what is linux and machine it is ubuntu / X86?

terryyizhong commented 5 years ago

It's ubuntu 16.04, with X86_64

terryyizhong commented 5 years ago

I am now trying build tensorflow with bazel. I first install bazel v0.24.0. But it shows I should downgrade to v0.21.0 to build tensorflow when I do ./configure. After I reinstall bazel v0.21. ./configure was success.

terryyizhong commented 5 years ago

The bazel build seems work now after I deactive my virtualenv. It is compling now though there are lots of warning.

ysujiang commented 4 years ago

在我停用virtualenv之后，Bazel构建现在似乎可以正常工作了。 尽管有很多警告，现在正在编译。

Have you got audio samples? Can you give me some? looking forward your audio samples.

alokprasad commented 4 years ago

@ysujiang i will upload sample soon , but did u trained the lpcnet also ?

ysujiang commented 4 years ago

@ysujiang i will upload sample soon , but did u trained the lpcnet also ?

I want to use the LPCNET as the vocoder ,but i haven't trained . Is it better than griffin_lim? can you send some samples synthesized by tacotron2+LPC model to my email [ ysujiangpei@163.com]? looking forward your audio samples.

ysujiang commented 4 years ago

@ysujiang i will upload sample soon , but did u trained the lpcnet also ?

@alokprasad hello，i trained the lpcnet ， i synthesized the samples by tacotron2+LPC model ，but the trill of speech is very serious，do you know why？and any changes to the model?