wassname/TTS
1
0
Fork 0
mirror of https://github.com/wassname/TTS.git synced 2026-09-10 11:50:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

pep8 format all

Browse Source
This commit is contained in:
Eren G
2018-08-02 16:34:17 +02:00
parent 3238ffa3e6
commit f5537dc48f
32 changed files with 766 additions and 599 deletions
Download Patch File Download Diff File Expand all files Collapse all files
server/README.md
+5 -5
View File
@@ -1,9 +1,9 @@
## TTS example web-server
Steps to run:
1. Download one of the models given on the main page.
2. Checkout the corresponding commit history.
2. Set paths and other options in the file ```server/conf.json```.
3. Run the server ```python server/server.py -c conf.json```. (Requires Flask)
1. Download one of the models given on the main page. Click [here](https://drive.google.com/drive/folders/1Q6BKeEkZyxSGsocK2p_mqgzLwlNvbHFJ?usp=sharing) for the lastest model.
2. Checkout the corresponding commit history or use ```server``` branch if you like to use the latest model.
2. Set the paths and the other options in the file ```server/conf.json```.
3. Run the server ```python server/server.py -c server/conf.json```. (Requires Flask)
4. Go to ```localhost:[given_port]``` and enjoy.
Note that the audio quality on browser is slightly worse due to the encoder quantization.
For high quality results, please use the library versions shown in the ```requirements.txt``` file.
server/conf.json
+1 -1
View File
@@ -1,5 +1,5 @@
{
"model_path":"/home/erogol/projects/models/LJSpeech/May-22-2018_03_24PM-e6112f7",
"model_path":"../models/May-22-2018_03_24PM-e6112f7",
"model_name":"checkpoint_272976.pth.tar",
"model_config":"config.json",
"port": 5002,
server/server.py
+8 -7
View File
@@ -2,12 +2,11 @@
import argparse
from synthesizer import Synthesizer
from TTS.utils.generic_utils import load_config
from flask import (Flask, Response, request,
render_template, send_file)
from flask import Flask, Response, request, render_template, send_file
parser = argparse.ArgumentParser()
parser.add_argument('-c', '--config_path', type=str,
help='path to config file for training')
parser.add_argument(
'-c', '--config_path', type=str, help='path to config file for training')
args = parser.parse_args()
config = load_config(args.config_path)
@@ -16,17 +15,19 @@ synthesizer = Synthesizer()
synthesizer.load_model(config.model_path, config.model_name,
config.model_config, config.use_cuda)
@app.route('/')
def index():
return render_template('index.html')
@app.route('/api/tts', methods=['GET'])
def tts():
text = request.args.get('text')
print(" > Model input: {}".format(text))
data = synthesizer.tts(text)
return send_file(data,
mimetype='audio/wav')
return send_file(data, mimetype='audio/wav')
if __name__ == '__main__':
app.run(debug=True, host='0.0.0.0', port=config.port)
app.run(debug=True, host='0.0.0.0', port=config.port)
server/synthesizer.py
+24 -18
View File
@@ -13,39 +13,44 @@ from matplotlib import pylab as plt
class Synthesizer(object):
def load_model(self, model_path, model_name, model_config, use_cuda):
model_config = os.path.join(model_path, model_config)
self.model_file = os.path.join(model_path, model_name)
self.model_file = os.path.join(model_path, model_name)
print(" > Loading model ...")
print(" | > model config: ", model_config)
print(" | > model file: ", self.model_file)
config = load_config(model_config)
self.config = config
self.use_cuda = use_cuda
self.model = Tacotron(config.embedding_size, config.num_freq, config.num_mels, config.r)
self.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=60)
self.model = Tacotron(config.embedding_size, config.num_freq,
config.num_mels, config.r)
self.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=60)
# load model state
if use_cuda:
cp = torch.load(self.model_file)
else:
cp = torch.load(self.model_file, map_location=lambda storage, loc: storage)
cp = torch.load(
self.model_file, map_location=lambda storage, loc: storage)
# load the model
self.model.load_state_dict(cp['model'])
if use_cuda:
self.model.cuda()
self.model.eval()
self.model.eval()
def save_wav(self, wav, path):
wav *= 32767 / max(1e-8, np.max(np.abs(wav)))
# sf.write(path, wav.astype(np.int32), self.config.sample_rate, format='wav')
# wav = librosa.util.normalize(wav.astype(np.float), norm=np.inf, axis=None)
# wav = wav / wav.max()
# sf.write(path, wav.astype('float'), self.config.sample_rate, format='ogg')
scipy.io.wavfile.write(path, self.config.sample_rate, wav.astype(np.int16))
# librosa.output.write_wav(path, wav.astype(np.int16), self.config.sample_rate, norm=True)
librosa.output.write_wav(path, wav.astype(np.int16),
self.config.sample_rate)
def tts(self, text):
text_cleaner = [self.config.text_cleaner]
@@ -54,14 +59,15 @@ class Synthesizer(object):
if len(sen) < 3:
continue
sen = sen.strip()
sen +='.'
sen += '.'
print(sen)
sen = sen.strip()
seq = np.array(text_to_sequence(text, text_cleaner))
chars_var = torch.from_numpy(seq).unsqueeze(0)
chars_var = torch.from_numpy(seq).unsqueeze(0).long()
if self.use_cuda:
chars_var = chars_var.cuda()
mel_out, linear_out, alignments, stop_tokens = self.model.forward(chars_var)
mel_out, linear_out, alignments, stop_tokens = self.model.forward(
chars_var)
linear_out = linear_out[0].data.cpu().numpy()
wav = self.ap.inv_spectrogram(linear_out.T)
# wav = wav[:self.ap.find_endpoint(wav)]