Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
coqui-tts/notebooks/Benchmark.ipynb

10 KiB
Raw Blame History

None <html lang="en"> <head> </head>
In [ ]:
%load_ext autoreload
%autoreload 2
import os
import sys
import io
import torch 
import time
import numpy as np
from collections import OrderedDict
from matplotlib import pylab as plt

%pylab inline
rcParams["figure.figsize"] = (16,5)
sys.path.append('/home/erogol/projects/')

import librosa
import librosa.display

from TTS.models.tacotron import Tacotron 
from TTS.layers import *
from TTS.utils.data import *
from TTS.utils.audio import AudioProcessor
from TTS.utils.generic_utils import load_config
from TTS.utils.text import text_to_sequence

import IPython
from IPython.display import Audio
from utils import *
In [ ]:
def tts(model, text, CONFIG, use_cuda, ap, figures=True):
    t_1 = time.time()
    waveform, alignment, spectrogram, stop_tokens = create_speech(model, text, CONFIG, use_cuda, ap) 
    print(" >  Run-time: {}".format(time.time() - t_1))
    if figures:                                                                                                         
        visualize(alignment, spectrogram, stop_tokens, CONFIG)                                                                       
    IPython.display.display(Audio(waveform, rate=CONFIG.sample_rate))  
    out_path = 'benchmark_samples/'
    os.makedirs(out_path, exist_ok=True)
    file_name = text.replace(" ", "_").replace(".","") + ".wav"
    out_path = os.path.join(out_path, file_name)
    ap.save_wav(waveform, out_path)
    return alignment, spectrogram, stop_tokens
In [ ]:
# Set constants
ROOT_PATH = '/data/shared/erogol_models/May-22-2018_03:24PM-loc-sen-attn-e6112f7/'
MODEL_PATH = ROOT_PATH + '/checkpoint_272976.pth.tar'
CONFIG_PATH = ROOT_PATH + '/config.json'
OUT_FOLDER = ROOT_PATH + '/test/'
CONFIG = load_config(CONFIG_PATH)
use_cuda = True
In [ ]:
# load the model
model = Tacotron(CONFIG.embedding_size, CONFIG.num_freq, CONFIG.num_mels, CONFIG.r)

# load the audio processor

ap = AudioProcessor(CONFIG.sample_rate, CONFIG.num_mels, CONFIG.min_level_db,
                    CONFIG.frame_shift_ms, CONFIG.frame_length_ms, CONFIG.preemphasis,
                    CONFIG.ref_level_db, CONFIG.num_freq, CONFIG.power, griffin_lim_iters=30)         


# load model state
if use_cuda:
    cp = torch.load(MODEL_PATH)
else:
    cp = torch.load(MODEL_PATH, map_location=lambda storage, loc: storage)

# load the model
model.load_state_dict(cp['model'])
if use_cuda:
    model.cuda()
model.eval()

EXAMPLES FROM TRAINING SET

In [ ]:
import pandas as pd
df = pd.read_csv('/data/shared/KeithIto/LJSpeech-1.0/metadata_val.csv', delimiter='|')
In [ ]:
sentence = df.iloc[175, 1]
print(sentence)
model.decoder.max_decoder_steps = 250
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence =  "It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent."
model.decoder.max_decoder_steps = 250
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap, figures=True)
In [ ]:
sentence = "Be a voice,not an echo."  # 'echo' is not in training set. 
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "The human voice is the most perfect instrument of all."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "I'm sorry Dave. I'm afraid I can't do that."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "This cake is great. It's so delicious and moist."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Generative adversarial network or variational auto-encoder."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Scientists at the CERN laboratory say they have discovered a new particle."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Here’s a way to measure the acute emotional intelligence that has never gone out of style."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "President Trump met with other leaders at the Group of 20 conference."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "The buses aren't the problem, they actually provide a solution."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Generative adversarial network or variational auto-encoder."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Basilar membrane and otolaryngology are not auto-correlations."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "He has read the whole thing."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "He reads books."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Thisss isrealy awhsome."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "This is your internet browser, Firefox."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "This is your internet browser Firefox."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "The quick brown fox jumps over the lazy dog."
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
sentence = "Does the quick brown fox jump over the lazy dog?"
align, spec, stop_tokens = tts(model, sentence, CONFIG, use_cuda, ap)
In [ ]:
!zip benchmark_samples/samples.zip benchmark_samples/*
</html>