add fallback sample and snr

app.py

@@ -177,7 +177,7 @@ def demo_fn(speech_upl: str, noise_type: str, snr: int, mic_input: Optional[str]
     snr = int(snr)
     noise_fn = NOISES[noise_type]
 
-    # 1. Load Clean Speech
+    # Load Clean Speech
     max_s = 10
     if speech_upl is not None:
         sample, _ = load_audio_torch(speech_upl, sr)
@@ -186,24 +186,23 @@ def demo_fn(speech_upl: str, noise_type: str, snr: int, mic_input: Optional[str]
             start = torch.randint(0, sample.shape[-1] - max_len, ()).item()
             sample = sample[..., start : start + max_len]
     else:
-        # Fallback sample
-        sample, _ = load_audio_torch("samples/p232_013_clean.wav", sr)
+        sample, _ = load_audio_torch("samples/fallback_sample.wav", sr)
         sample = sample[..., : max_s * sr]
 
     # Ensure channels first
     if sample.dim() > 1 and sample.shape[0] > 1:
         sample = sample.mean(dim=0, keepdim=True)
 
-    # 2. Add Noise (if selected)
+    # Add Noise (if selected)
     if noise_fn is not None:
         noise, _ = load_audio_torch(noise_fn, sr)
         _, _, sample = mix_at_snr(sample, noise, snr)
 
-    # 3. Save Noisy File (Input for enhance_audio)
+    # Save Noisy File (Input for enhance_audio)
     noisy_wav_path = tempfile.NamedTemporaryFile(suffix="noisy.wav", delete=False).name
     save_audio_torch(noisy_wav_path, sample, sr)
 
-    # 4. Run Inference using your Custom Function
+    # Run Inference using your Custom Function
     enhanced_wav_path = tempfile.NamedTemporaryFile(suffix="enhanced.wav", delete=False).name
 
     logger.info("Starting enhancement...")
@@ -212,10 +211,8 @@ def demo_fn(speech_upl: str, noise_type: str, snr: int, mic_input: Optional[str]
     enhance_audio_hf(None, noisy_wav_path, enhanced_wav_path)
     logger.info("Enhancement finished")
 
-    # 5. Load Enhanced Audio for Visualization
     enhanced, _ = load_audio_torch(enhanced_wav_path, sr)
 
-    # 6. Generate Visuals
     ax_noisy.clear()
     ax_enh.clear()
     noisy_im = spec_im(sample, sr=sr, figure=fig_noisy, ax=ax_noisy)
@@ -248,16 +245,16 @@ with gr.Blocks() as demo:
             audio_file = gr.Audio(type="filepath", label="File Input", visible=True)
             inputs = [
                 audio_file,
-                # gr.Dropdown(
-                #     label="Add background noise",
-                #     choices=list(NOISES.keys()),
-                #     value="None",
-                # ),
-                # gr.Dropdown(
-                #     label="Noise Level (SNR)",
-                #     choices=["-5", "0", "10", "20"],
-                #     value="10",
-                # ),
+                gr.Dropdown(
+                    label="Add background noise",
+                    choices=list(NOISES.keys()),
+                    value="None",
+                ),
+                gr.Dropdown(
+                    label="Noise Level (SNR)",
+                    choices=["-5", "0", "10", "20"],
+                    value="10",
+                ),
                 mic_input,
             ]
             btn = gr.Button("Denoise", variant="primary")

samples/fallback_sample.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:572cb9eef7bf796a102c5d82d09e592b15ac97d22daf15c54f7b5a91139720a8
+size 320058