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

Update vits dataset

This commit is contained in:
Eren Gölge 2022-03-02 09:14:32 +01:00
parent a84499c5da
commit 942df0fb05
2 changed files with 6 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
TTS/tts/models/vits.py
View File

@ -196,14 +196,12 @@ class VitsDataset(TTSDataset):
def __getitem__(self, idx): def __getitem__(self, idx):
item = self.samples[idx] item = self.samples[idx]
raw_text = item["text"]
text, wav_file, speaker_name, language_name, _ = _parse_sample(item) wav, _ = load_audio(item["audio_file"])
raw_text = text wav_filename = os.path.basename(item["audio_file"])
wav, _ = load_audio(wav_file) token_ids = self.get_token_ids(idx, item["text"])
wav_filename = os.path.basename(wav_file)
token_ids = self.get_token_ids(idx, text)
# after phonemization the text length may change # after phonemization the text length may change
# this is a shameful 🤭 hack to prevent longer phonemes # this is a shameful 🤭 hack to prevent longer phonemes
@ -218,8 +216,8 @@ class VitsDataset(TTSDataset):
"token_len": len(token_ids), "token_len": len(token_ids),
"wav": wav, "wav": wav,
"wav_file": wav_filename, "wav_file": wav_filename,
"speaker_name": speaker_name, "speaker_name": item["speaker_name"],
"language_name": language_name, "language_name": item["language"],
} }
@property @property

20
TTS/utils/synthesizer.py
View File

@ -126,26 +126,6 @@ class Synthesizer(object):
self.encoder_checkpoint = self.tts_config.model_args.speaker_encoder_model_path self.encoder_checkpoint = self.tts_config.model_args.speaker_encoder_model_path
self.encoder_config = self.tts_config.model_args.speaker_encoder_config_path self.encoder_config = self.tts_config.model_args.speaker_encoder_config_path
def _is_use_speaker_embedding(self):
"""Check if the speaker embedding is used in the model"""
# we handle here the case that some models use model_args some don't
use_speaker_embedding = False
if hasattr(self.tts_config, "model_args"):
use_speaker_embedding = self.tts_config["model_args"].get("use_speaker_embedding", False)
use_speaker_embedding = use_speaker_embedding or self.tts_config.get("use_speaker_embedding", False)
return use_speaker_embedding
def _is_use_d_vector_file(self):
"""Check if the d-vector file is used in the model"""
# we handle here the case that some models use model_args some don't
use_d_vector_file = False
if hasattr(self.tts_config, "model_args"):
config = self.tts_config.model_args
use_d_vector_file = config.get("use_d_vector_file", False)
config = self.tts_config
use_d_vector_file = use_d_vector_file or config.get("use_d_vector_file", False)
return use_d_vector_file
def _load_vocoder(self, model_file: str, model_config: str, use_cuda: bool) -> None: def _load_vocoder(self, model_file: str, model_config: str, use_cuda: bool) -> None:
"""Load the vocoder model. """Load the vocoder model.