Update app.py

app.py

@@ -20,13 +20,16 @@ def generate_text(audio):
         return "No audio received"
     
     sr, data = audio  # Unpack the tuple
-    print(sr, data.dtype)
-    
-    data = data.astype(np.float32) / 32768.0
-    data = librosa.resample(data, orig_sr=sr, target_sr=16000)
-    sr = 16000
-   
-    print(sr, data.dtype)
+    print(f"Original sample rate: {sr}, dtype: {data.dtype}")
+
+    # Convert to float32 and ensure 16kHz
+    data = data.astype(np.float32)
+    if sr != 16000:
+        data = librosa.resample(data, orig_sr=sr, target_sr=16000)
+        sr = 16000
+
+    print(f"Processed sample rate: {sr}, dtype: {data.dtype}")
+
     inputs = asr_processor(data, sampling_rate=sr, return_tensors="pt", padding=True)
     
     with torch.no_grad():
@@ -37,15 +40,13 @@ def generate_text(audio):
     return asr_processor.decode(pred_ids)
 
 
+# Ensure Gradio records at 16kHz and float32
 input_audio = gr.Audio(
     sources=["microphone"],
-    waveform_options=gr.WaveformOptions(
-        waveform_color="#01C6FF",
-        waveform_progress_color="#0066B4",
-        skip_length=2,
-        show_controls=False,
-    ),
+    type="numpy",  # Ensures we receive (sr, np.ndarray)
+    sample_rate=16000,  # Force 16kHz recording
 )
+
 demo = gr.Interface(
     fn=generate_text,
     inputs=input_audio,