Spaces:
Running on Zero
Running on Zero
| """ | |
| vad.py — Detector de voz con red neuronal (Silero VAD v5, ONNX, licencia MIT). | |
| A diferencia del detector por energía, distingue la voz de otros sonidos fuertes (un cláxon, una | |
| puerta, un golpe): sirve para encontrar ruidos sueltos en las pausas sin tocar la voz. | |
| Trabaja a 16 kHz en ventanas de 512 muestras (32 ms). | |
| """ | |
| from __future__ import annotations | |
| import os | |
| import threading | |
| import numpy as np | |
| SR = 16000 | |
| WIN = 512 | |
| CONTEXT = 64 | |
| HOP_S = WIN / SR | |
| _lock = threading.Lock() | |
| _session = None | |
| def _load(): | |
| global _session | |
| with _lock: | |
| if _session is None: | |
| import onnxruntime as ort | |
| from models import silero_vad_path | |
| opts = ort.SessionOptions() | |
| opts.log_severity_level = 3 | |
| opts.intra_op_num_threads = 1 # modelo diminuto: un hilo es lo más rápido | |
| opts.inter_op_num_threads = 1 | |
| _session = ort.InferenceSession(str(silero_vad_path()), opts, providers=["CPUExecutionProvider"]) | |
| return _session | |
| def speech_prob(mono: np.ndarray, sr: int) -> np.ndarray: | |
| """Probabilidad de voz (0–1) por ventana de 32 ms.""" | |
| import soxr | |
| sess = _load() | |
| x = soxr.resample(np.asarray(mono, dtype=np.float32), sr, SR, quality="HQ") if sr != SR else mono | |
| x = np.asarray(x, dtype=np.float32) | |
| n = int(np.ceil(len(x) / WIN)) | |
| x = np.pad(x, (0, n * WIN - len(x))) | |
| state = np.zeros((2, 1, 128), dtype=np.float32) | |
| ctx = np.zeros(CONTEXT, dtype=np.float32) | |
| sr_arr = np.array(SR, dtype=np.int64) | |
| out = np.empty(n, dtype=np.float32) | |
| for i in range(n): | |
| chunk = x[i * WIN : (i + 1) * WIN] | |
| inp = np.concatenate([ctx, chunk])[None, :] | |
| p, state = sess.run(None, {"input": inp, "state": state, "sr": sr_arr}) | |
| out[i] = p[0, 0] | |
| ctx = chunk[-CONTEXT:] | |
| return out | |
| def available() -> bool: | |
| try: | |
| _load() | |
| return True | |
| except Exception as e: | |
| print(f"[VAD] No disponible: {e}") | |
| return False | |