github/ebook2audiobook
1
1
Fork 0
mirror of https://github.com/DrewThomasson/ebook2audiobook.git synced 2026-01-09 22:08:13 -05:00
Code Issues Packages Projects Releases Wiki Activity

v25.12.28

Browse Source
This commit is contained in:
ROBERT MCDOWELL
2025-12-28 17:56:29 -08:00
committed by GitHub
parent 7595d1a4e4 ce4ed79940
commit 927ea76b4a
1 changed files with 4 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
lib/classes/tts_engines/common/utils.py
View File

@@ -214,16 +214,15 @@ class TTSUtils:
audio_tensor = audio_sentence.detach().cpu().unsqueeze(0) audio_tensor = audio_sentence.detach().cpu().unsqueeze(0)
elif isinstance(audio_sentence, np.ndarray): elif isinstance(audio_sentence, np.ndarray):
audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0) audio_tensor = torch.from_numpy(audio_sentence).unsqueeze(0)
audio_tensor = audio_tensor.cpu()
elif isinstance(audio_sentence, (list, tuple)): elif isinstance(audio_sentence, (list, tuple)):
audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0) audio_tensor = torch.tensor(audio_sentence, dtype=torch.float32).unsqueeze(0)
audio_tensor = audio_tensor.cpu()
else: else:
error = f"Unsupported XTTSv2 wav type: {type(audio_sentence)}" error = f"{self.session['tts_engine']}: Unsupported wav type: {type(audio_sentence)}"
print(error) print(error)
return False return False
if audio_sentence is not None: if audio_tensor is not None and audio_tensor.numel() > 0:
audio_sentence = audio_sentence.tolist()
sourceTensor = self._tensor_type(audio_sentence)
audio_tensor = sourceTensor.clone().detach().unsqueeze(0).cpu()
# CON is a reserved name on windows # CON is a reserved name on windows
lang_dir = 'con-' if self.session['language'] == 'con' else self.session['language'] lang_dir = 'con-' if self.session['language'] == 'con' else self.session['language']
new_voice_path = re.sub(r'([\\/])eng([\\/])', rf'\1{lang_dir}\2', voice_path) new_voice_path = re.sub(r'([\\/])eng([\\/])', rf'\1{lang_dir}\2', voice_path)