update

app.py

@@ -95,7 +95,7 @@ def check_audio_length(audio_path, max_duration=MAX_AUDIO_DURATION):
         # If we can't determine the length, we'll try to process it anyway
         return True, 0
 
-@spaces.GPU
+@spaces.GPU(duration=120)
 def process_with_progress(
     progress=gr.Progress(),
     input_audio=None,

infer.py

@@ -416,11 +416,11 @@ def process_segment(
 @click.option('--fade-duration', type=float, default=20.0, help='Fade duration in milliseconds')
 @click.option('--sliced-inference', is_flag=True, default=False, help='Use sliced inference for processing long segments')
 @click.option('--robust-f0', type=int, default=0, help='Level of robust f0 filtering (0=none, 1=light, 2=aggressive)')
-@click.option('--slicer-threshold', type=float, default=-35.0, help='Threshold for audio slicing in dB')
+@click.option('--slicer-threshold', type=float, default=-30.0, help='Threshold for audio slicing in dB')
 @click.option('--slicer-min-length', type=int, default=3000, help='Minimum length of audio segments in milliseconds')
 @click.option('--slicer-min-interval', type=int, default=100, help='Minimum interval between audio segments in milliseconds')
 @click.option('--slicer-hop-size', type=int, default=10, help='Hop size for audio slicing in milliseconds')
-@click.option('--slicer-max-sil-kept', type=int, default=300, help='Maximum silence kept in milliseconds')
+@click.option('--slicer-max-sil-kept', type=int, default=200, help='Maximum silence kept in milliseconds')
 @click.option('--use-fp16', is_flag=True, default=True, help='Use float16 precision for faster inference')
 def main(
     model,