| 379 | return human_speech_array |
| 380 | |
| 381 | def process_tts_single(text, save_dir, voice1): |
| 382 | s1_sentences = [] |
| 383 | |
| 384 | pipeline = KPipeline(lang_code='a', repo_id='weights/Kokoro-82M') |
| 385 | |
| 386 | voice_tensor = torch.load(voice1, weights_only=True) |
| 387 | generator = pipeline( |
| 388 | text, voice=voice_tensor, # <= change voice here |
| 389 | speed=1, split_pattern=r'\n+' |
| 390 | ) |
| 391 | audios = [] |
| 392 | for i, (gs, ps, audio) in enumerate(generator): |
| 393 | audios.append(audio) |
| 394 | audios = torch.concat(audios, dim=0) |
| 395 | s1_sentences.append(audios) |
| 396 | s1_sentences = torch.concat(s1_sentences, dim=0) |
| 397 | save_path1 =f'{save_dir}/s1.wav' |
| 398 | sf.write(save_path1, s1_sentences, 24000) # save each audio file |
| 399 | s1, _ = librosa.load(save_path1, sr=16000) |
| 400 | return s1, save_path1 |
| 401 | |
| 402 | |
| 403 | |