Perform TTS inference and save the generated audio.
(
text,
model,
prompt_text=None,
prompt_speech=None,
gender=None,
pitch=None,
speed=None,
save_dir="example/results",
)
| 49 | |
| 50 | |
| 51 | def run_tts( |
| 52 | text, |
| 53 | model, |
| 54 | prompt_text=None, |
| 55 | prompt_speech=None, |
| 56 | gender=None, |
| 57 | pitch=None, |
| 58 | speed=None, |
| 59 | save_dir="example/results", |
| 60 | ): |
| 61 | """Perform TTS inference and save the generated audio.""" |
| 62 | logging.info(f"Saving audio to: {save_dir}") |
| 63 | |
| 64 | if prompt_text is not None: |
| 65 | prompt_text = None if len(prompt_text) <= 1 else prompt_text |
| 66 | |
| 67 | # Ensure the save directory exists |
| 68 | os.makedirs(save_dir, exist_ok=True) |
| 69 | |
| 70 | # Generate unique filename using timestamp |
| 71 | timestamp = datetime.now().strftime("%Y%m%d%H%M%S") |
| 72 | save_path = os.path.join(save_dir, f"{timestamp}.wav") |
| 73 | |
| 74 | logging.info("Starting inference...") |
| 75 | |
| 76 | # Perform inference and save the output audio |
| 77 | with torch.no_grad(): |
| 78 | wav = model.inference( |
| 79 | text, |
| 80 | prompt_speech, |
| 81 | prompt_text, |
| 82 | gender, |
| 83 | pitch, |
| 84 | speed, |
| 85 | ) |
| 86 | |
| 87 | sf.write(save_path, wav, samplerate=16000) |
| 88 | |
| 89 | logging.info(f"Audio saved at: {save_path}") |
| 90 | |
| 91 | return save_path |
| 92 | |
| 93 | |
| 94 | def build_ui(model_dir, device=0): |
no test coverage detected