Closed nkeilar closed 3 weeks ago
Not at all, you certainly can. Use --xtts_device cpu
if you want to force loading xtts onto CPU. If you don't have a GPU, this should happen automatically.
If using the docker, just set this in your speech.env:
EXTRA_ARGS=--xtts_device cpu
I noted that CPU only support seems possible:
https://docs.coqui.ai/en/latest/docker_images.html
Is there any technical reason we could not utilize this, even if slower, on CPU?