Missing referenced Norwegian voice "no_NO-nvcc-medium"
The Home Assistant Piper app just got updated and the changelog lists "Add 1 new Norwegian voice (no_NO-nvcc-medium)", but trying to use it results in wyoming_piper.download.VoiceNotFoundError: no_NO-nvcc-medium as the voice model files do not exist here.
I've scoured here and other repos without finding any sort of PR or reference to these files actually existing, even as it was added to voices.json here in january.
Any ideas what happened here? Would love this voice model to work!
I did train this voice, but it ended up producing garbled output. I must have accidentally committed the metadata 😬
I put it in the queue to retrain; hopefully it will produce something better this time around.
Ah, figured it was something like that. Fingers crossed, got excited for a good multi-speaker Norwegian model!
It sounds better than before, but I'm not very happy with it: https://rhasspy.github.io/piper-samples/#no_NO-nvcc-medium
Speaker #5 (MSV) seems to sound the best to me, but I only speak English.
Thanks for the effort! As a native speaker I can say it's far from the quality of the previous model. Very mumbly and jumpy. It sounds like multiple dialects are mixed in, which could be a big part of why it sounds strange. Norwegian dialects have drastically different intonation. I haven't looked into it, but can the dataset be split by dialects in any way? Or is each speaker in this model already based on recordings of a single person?