main.py

# V3
import os
import torch
import torchaudio

device = torch.device('cpu')
torch.set_num_threads(4)
local_file = 'model.pt'

if not os.path.isfile(local_file):
    torch.hub.download_url_to_file(
        'https://models.silero.ai/models/tts/tt/v3_tt.pt', local_file)

model = torch.package.PackageImporter(
    local_file).load_pickle("tts_models", "model")
model.to(device)

example_text = 'Мин почти кеше тавышы. Мин сезнең белән д+ус булырга телим'

# җ
# ү
# ә
# һ
# ң
# ө

text = example_text
filename = text+'.mp3'

sample_rate = 48000
speaker = 'dilyara'
put_accent = True

audio_paths = model.apply_tts(text=text+'.',
                              speaker=speaker,
                              sample_rate=sample_rate,
                              put_accent=put_accent,
                              )


torchaudio.save('audio/' + filename,
                audio_paths.unsqueeze(0),
                sample_rate=sample_rate)