Voice files in Safetensors format?

[MackNcD]

I'd love to use Piper for my program but there are no safetensor models for the voice models, so I can't guarantee people will be safe to use it. Can you guys convert the models to safetensors? Or will the program at least read them if I convert myself?

[nmstoker]

Which part(s) of piper do you think is/are unsafe?

Looking at the README for safetensor they highlight that they provide safety by avoiding pickle, yet pickle isn't used in this repo, it uses onnx. The onnx/protobuf entry in the table in the safetensor repo marks it as safe so I'm not convinced of the validity of your concern, but do say if there's a particular aspect I'm overlooking.