enSongDec Neural-to-AudioEmbeddings-to-AudioWaveform model using EnCodec for neurally-driven vocal prostheses