I am trying to setup the rhasspy/wyoming-piper container on my beefy server so my homeassistant will reach out to it for TTS. How every i cant start the container with a custom model. It always throws an error that i couldt find the model on higging face
#Cutsom Piper Model in docker container
grim lark
supple heron
I am not sure how to do that on a remote server. I am thinking that you might have to point to a local copy on the server because that is what you have to do to install a custom model to the HA piper addon. The .json and .onnx files of the model have to be stored in the share/piper directory for the HA addon. I went back to using the HA piper addon so I did not try to figure it out. I am running HA and the piper addon on a N100 mini computer. I don't really notice the difference in speed from piper running remotely on an old Intel I7 server and the piper addon for HA even for high quality models. Neither piper instance is using the GPU. Of course the speed difference is significant between the remote server version of wyoming/whisper using the GPU and the HA whisper addon sans GPU.
grim lark
Thank you, i guess i will keep using the HA Addon
silver meadow
are you using an onnx file?
If you look into https://github.com/rhasspy/wyoming-piper/blob/master/wyoming_piper/__main__.py it kinda shows how it would work
Seems that you would drop your onnx file into the data dir (same one piper downloads voices to from huggingface) and if it is there, it uses it
modern cloak
I am trying to setup the rhasspy/wyoming-piper container on my beefy server so m...
Of course you can run piper container on another machine.
But I cant help with your error message. Sounds like you didn’t download a model, but you’re trying to point to a remote source 🤷♂️
finite arch