Add F5 TTS

[pabloortegaa]

I just saw this TTS model(https://github.com/SWivid/F5-TTS), which works very well for English. Are you planning on including it on the project? Thanks!!

[KoljaB]

Absolutely thinking about that, I love that TTS system. Things that keep me off currently:

1. Not on pypi - without pip install I need to copy their whole repo, don't want that
2. No streaming support so bigger latency than for other engines
   Both are kind of bummers but 1 is the bigger one. I need a solution to gracefully install it with RealtimeTTS

[aadr1024]

Hi! @KoljaB, would pip install git+https://github.com/SWivid/F5-TTS.git not suffice as a pip install?

[KoljaB]

Sadly not, pypi would not accept this within a setup file. So I can't integrate it to be installed with "pip install RealtimeTTS[F5]" currently. Also no streaming support is still a bummer.
Still my first pick TTS system that I would love to integrate (together with GPT-SoVITS).

[KoljaB]

Looked into F5-TTS code.

Implementing real-time streaming is not trivial because the model processes entire sequences at once using full-sequence attention and ODE integration. This doesn't support incremental output. So it would require significant architectural changes to support incremental, real-time computation needed for streaming.