Support offline mobile - Githubissues

coqui-ai / TTS

🐸💬 - a deep learning toolkit for Text-to-Speech, battle-tested in research and production

http://coqui.ai

Mozilla Public License 2.0

34.62k stars 4.2k forks source link

Closed boxabirds closed 3 years ago

boxabirds commented 3 years ago

This repository is really great! Decent samples too.

I see a huge opportunity for this to be extended to support mobile.

There are a number of obstacles to this of course, including running on TF-Lite.

If you ported it to Dart you could transpile it to iOS and Android.

erogol commented 3 years ago

Thx ! We also plan to expand :frog: TTS's use in offline and low-compute devices.

Currently, some models we release are already real-time using simply the torch back-end.

Targeting the mobile, we can make use of Torch Script, ONNX or TF-Lite export. So one step at a time we plan to get there.

Actually you can already run the vanilla Tacotron on TF-Lite. A bit of documentation is here https://github.com/coqui-ai/TTS/wiki/Converting-Torch-Tacotron-to-TF-2.0

Dmole commented 3 years ago

There are currently 3 FOSS TTSEs for Mobile https://search.f-droid.org/?q=tts+engine&lang=en A coqui-ai.apk could improve that short list.