PABannier / bark.cpp

Suno AI's Bark model in C/C++ for fast text-to-speech
MIT License
630 stars 48 forks source link

TTS inference using bark.cpp doesn't work properly on Android phone #178

Open zhouwg opened 1 month ago

zhouwg commented 1 month ago

The Project bark.cpp is excellent and thanks for your amazing bark.cpp. I want to utilize the bark.cpp as TTS engine in my learning project.

TTS inference using bark.cpp doesn't work properly on Android phone currently https://github.com/zhouwg/kantv/blob/master/core/ggml/barkcpp/bark.cpp#L1347

https://github.com/zhouwg/kantv/blob/master/core/ggml/barkcpp/bark.cpp#L1558

https://github.com/zhouwg/kantv/blob/master/core/ggml/barkcpp/bark.cpp#L1751

I found that GGML's internal / GGML API changes quickly/frequently:

https://github.com/zhouwg/kantv/blob/master/core/ggml/llamacpp/ggml.c#L4696

https://github.com/zhouwg/kantv/blob/master/core/ggml/llamacpp/ggml.c#L5848

Could you help to fix this issue(merge upstream GGML source code to your encodec.cpp because you are the original author of bark.cpp and know everything/details in internal of bark.cpp)?

thanks for your time and thanks so much.

PABannier commented 4 weeks ago

Thanks @zhouwg for the issue! I'm in the process of pulling the latest version of ggml and encodec.cpp into bark.cpp, but it is a tedious and painfully long. I'll ping you when it is done.

zhouwg commented 3 weeks ago

Thanks @zhouwg for the issue! I'm in the process of pulling the latest version of ggml and encodec.cpp into bark.cpp, but it is a tedious and painfully long. I'll ping you when it is done.

thanks for your feedback so much.

I had been tried(bark.cpp using the latest ggml) in my personal learning project but the running result is not correct(I know very little about real/hardcore AI tech so I don't know how to fix it): 467524614