Support of the XTTS 2 in tts-server

[jbgi]

Fix #99.

Based on original PR by @bivashy: coqui-ai#3476

[eginhard]

Cool, thanks! I'll test it.

[mikejgray]

FWIW, I built and tested this locally and have a Docker image at mjgray84/coqui:latest. Works great!

[eginhard]

Hey @jbgi ! Thanks again for opening this. I looked into it for a bit and think that instead of this it would be best to adapt the server to use Coqui's Python API internally (which wasn't available at the time the server was originally created). This means the server can support all current and future models without further adaptation.

If you, or anyone else, would like to do this, it should be fairly easy to do. It would be very similar to how the command line interface is implemented in TTS/bin/synthesize.py, i.e. creating a TTS instead of a Synthesizer instance. Feel free to ask me any questions. Otherwise I can take care of it eventually.

[jbgi]

If you, or anyone else, would like to do this, it should be fairly easy to do.

I'll leave it to someone else as I won't have the time to look into tts internals any time soon.

[eginhard]

I looked into it for a bit and think that instead of this it would be best to adapt the server to use Coqui's Python API internally (which wasn't available at the time the server was originally created). This means the server can support all current and future models without further adaptation.

Just did that in #252, so closing this PR.