Update denoiser.py

denoiser.py

@@ -211,13 +211,13 @@ class Denoiser:
         except Exception as e:
             logger.warning(f"[Denoiser] DeepFilterNet unavailable ({e})")
 
-        # ── Try RNNoise (best for voice, no musical artifacts) ───────
+        # ── Try Wiener filter (scipy — artifact-free, no compilation needed) ─
         try:
             result = self._rnnoise(audio, sr)
-            print("[Denoiser] ✅ RNNoise removal done")
-            return result, "RNNoise"
+            print("[Denoiser] ✅ Wiener filter noise removal done")
+            return result, "Wiener filter"
         except Exception as e:
-            logger.warning(f"[Denoiser] RNNoise unavailable ({e})")
+            logger.warning(f"[Denoiser] Wiener filter failed ({e})")
 
         # ── Fallback: noisereduce with mild settings ──────────────────
         # IMPORTANT: Keep prop_decrease LOW (0.50-0.60) to avoid musical
@@ -242,16 +242,20 @@ class Denoiser:
 
     def _rnnoise(self, audio, sr):
         """
-        RNNoise — Mozilla's RNN-based noise suppressor.
-        Specifically trained on voice, produces zero musical artifacts.
-        Requires: pip install rnnoise-python
+        Wiener filter via scipy — no compilation needed, works on HF Spaces.
+        Much cleaner than noisereduce for voice — no musical artifacts.
         """
-        import rnnoise
-        # RNNoise requires 48kHz mono float32
-        audio_48k = self._resample(audio, sr, 48000) if sr != 48000 else audio
-        denoised  = rnnoise.reduce_noise(audio_48k, 48000)
-        # Resample back to original SR
-        result = self._resample(denoised, 48000, sr) if sr != 48000 else denoised
+        from scipy.signal import wiener
+        # Wiener filter works best on short frames
+        frame_size = int(sr * 0.02)  # 20ms frames
+        result = np.zeros_like(audio)
+        for i in range(0, len(audio) - frame_size, frame_size):
+            frame = audio[i:i + frame_size]
+            result[i:i + frame_size] = wiener(frame, mysize=7)
+        # Handle last chunk
+        remainder = len(audio) % frame_size
+        if remainder:
+            result[-remainder:] = wiener(audio[-remainder:], mysize=7)
         return result.astype(np.float32)
 
     def _deepfilter(self, audio, sr):