Load piper into RAM/VRAM for persistence (remove model load time)

JohnnySn0w / Echo

Voice-to-voice personal assistant, Full-local, GPU company agnostic.

5 stars 1 forks source link

Load piper into RAM/VRAM for persistence (remove model load time) #1

Closed JohnnySn0w closed 3 months ago

JohnnySn0w commented 5 months ago

Depends on #2 Piper currently has to be loaded per invocation. It would be preferable to load it into ram and submit requests to it via http or other messaging protocol, to make invocation easier, and to save on load times.