malarinv
/
tacotron2
mirror of https://github.com/malarinv/tacotron2
1
0
Fork 0
Code Issues Releases Wiki Activity

Compare commits

base: malarinv:5a30069f0a25375d79f0bfd52fe39a07cd161a2c
Branches Tags
malarinv:master
malarinv:experiments
malarinv:clean_slate
malarinv:latest_model
malarinv:attention_full_mel
malarinv:fp16_path
malarinv:load_mel_from_disk
malarinv:padding-patch-0.4
malarinv:mask-utils-0.4
malarinv:single-gpu-and-0.4
..
compare: malarinv:dde35048b76662bd46b6d7b3cb31dede1afcd5f1
Branches Tags
malarinv:master
malarinv:experiments
malarinv:clean_slate
malarinv:latest_model
malarinv:attention_full_mel
malarinv:fp16_path
malarinv:load_mel_from_disk
malarinv:padding-patch-0.4
malarinv:mask-utils-0.4
malarinv:single-gpu-and-0.4

No commits in common. "5a30069f0a25375d79f0bfd52fe39a07cd161a2c" and "dde35048b76662bd46b6d7b3cb31dede1afcd5f1" have entirely different histories.
5a30069f0a ... dde35048b7

1 changed files with 7 additions and 26 deletions
Show Stats Expand all files Collapse all files

33
taco2/tts.py
View File

@ -5,8 +5,6 @@ import numpy as np
import torch import torch
import pyaudio import pyaudio
import klepto import klepto
import argparse
from pathlib import Path
from .model import Tacotron2 from .model import Tacotron2
from glow import WaveGlow from glow import WaveGlow
from .hparams import HParams from .hparams import HParams
@ -16,8 +14,7 @@ from .denoiser import Denoiser
from .audio_processing import griffin_lim, postprocess_audio from .audio_processing import griffin_lim, postprocess_audio
TTS_SAMPLE_RATE = 22050 TTS_SAMPLE_RATE = 22050
OUTPUT_SAMPLE_RATE = 22050 OUTPUT_SAMPLE_RATE = 16000
# OUTPUT_SAMPLE_RATE = 16000
# config from # config from
# https://github.com/NVIDIA/waveglow/blob/master/config.json # https://github.com/NVIDIA/waveglow/blob/master/config.json
@ -119,7 +116,7 @@ class TTSModel(object):
audio = audio.cpu().numpy() audio = audio.cpu().numpy()
return postprocess_audio( return postprocess_audio(
audio, tempo=0.6, src_rate=TTS_SAMPLE_RATE, dst_rate=OUTPUT_SAMPLE_RATE audio, src_rate=TTS_SAMPLE_RATE, dst_rate=OUTPUT_SAMPLE_RATE
) )
@ -136,7 +133,10 @@ def player_gen():
return play_device return play_device
def repl(tts_model): def repl():
tts_model = TTSModel(
"/Users/malar/Work/tacotron2_r4_83000.pt", "/Users/malar/Work/waveglow_484000"
)
player = player_gen() player = player_gen()
def loop(): def loop():
@ -148,26 +148,7 @@ def repl(tts_model):
def main(): def main():
parser = argparse.ArgumentParser( interactive_loop = repl()
formatter_class=argparse.ArgumentDefaultsHelpFormatter
)
parser.add_argument(
"-t",
"--tacotron2_path",
type=Path,
default="./tacotron.pt",
help="Path to a tacotron2 model",
)
parser.add_argument(
"-w",
"--waveglow_path",
type=Path,
default="./waveglow_256channels.pt",
help="Path to a waveglow model",
)
args = parser.parse_args()
tts_model = TTSModel(**vars(args))
interactive_loop = repl(tts_model)
while True: while True:
interactive_loop() interactive_loop()