adjust input sample rate to 16k and optimize chunking to 55.12s segments

app.py

@@ -35,23 +35,24 @@ def process_fn(input_audio_path, seed, guidance_scale, num_inference_steps):
         output_audio_path (str): the filepath of the processed audio.
     """
 
-    sig = AudioSignal(input_audio_path, sample_rate=48000)    
+    sig = AudioSignal(input_audio_path, sample_rate=16000)    
 
     outfile = "./output.wav"
 
     audio_concat = None
 
     total_length = sig.duration
-    num_segs = int(total_length / 10) #10 second segments
-    remainder = total_length % 10 # duration of last segment
+    num_segs = int(total_length / 5.12) # 5.12 second segments
+    remainder = total_length % 5.12 # duration of last segment
 
     for audio_segment in range(num_segs):
-        start = audio_segment * 10
+        print(f"Processing segment {audio_segment} of {num_segs}")
+        start = audio_segment * 5.12
 
         if audio_segment == num_segs - 1:
             end = start + remainder
         else:
-            end = start + 10
+            end = start + 5.12
 
         # get segment of audio from original file
         sig_seg = sig[start*sig.sample_rate:int(end*sig.sample_rate)]  # int accounts for float end time on last seg
@@ -110,4 +111,4 @@ with gr.Blocks() as webapp:
     ctrls_data, ctrls_button, process_button, cancel_button = build_endpoint(inputs, output, process_fn, card)
 
 #webapp.queue()
-webapp.launch(share=True)
+webapp.launch()