soundfile.write("out.wav", speech.numpy(), text2speech.fs, "PCM_16") gives an error about espnet_model_zoo

[MikeyBeez]

Sorry. I'm messing this up. I got an error running this interactively. Also, it create a wave file when run as a .py file, but the wav file only had a single beep. Here's the file:

import soundfile
from espnet_model_zoo.downloader import ModelDownloader
from espnet2.bin.tts_inference import Text2Speech
d = ModelDownloader()
text2speech = Text2Speech(**d.download_and_unpack("kan-bayashi/csmsc_conformer_fastspeech2"))

speech, *_ = text2speech("This is a test.")
soundfile.write("out.wav", speech.numpy(), text2speech.fs, "PCM_16")

Also, how do a use a model for TTS that I've already downloaded?

[kan-bayashi]

csmsc is mandarin model.
If you want to try English, use ljspeech model.

Also, how do a use a model for TTS that I've already downloaded?

See the returned values of
d.download_and_unpack("kan-bayashi/csmsc_conformer_fastspeech2")
You can understand.