theNeofr commited on
Commit
4788471
1 Parent(s): ac73748

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +13 -7
app.py CHANGED
@@ -338,7 +338,7 @@ def update_stems(model):
338
 
339
 
340
 
341
- with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base(primary_hue=gr.themes.colors.red, secondary_hue=gr.themes.colors.pink) as app:
342
  gr.HTML("<h1> 🎵 Audio-Separator 🎵 </h1>")
343
 
344
  with gr.Tab("Roformer"):
@@ -346,10 +346,12 @@ with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base(primary_hu
346
  with gr.Row():
347
  roformer_model = gr.Dropdown(label="Select the Model", choices=list(ROFORMER_MODELS.keys()))
348
  with gr.Row():
349
- roformer_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
350
- roformer_override_seg_size = gr.Checkbox(value=False, label="Override segment size", info="Override model default segment size instead of using the model default value.")
351
- roformer_overlap = gr.Slider(minimum=2, maximum=10, step=1, value=8, label="Overlap", info="Amount of overlap between prediction windows. Lower is better but slower.")
352
- roformer_pitch_shift = gr.Slider(minimum=-12, maximum=12, step=1, value=0, label="Pitch shift", info="Shift audio pitch by a number of semitones while processing. may improve output for deep/high vocals.")
 
 
353
  with gr.Row():
354
  roformer_audio = gr.Audio(label="Input Audio", type="filepath")
355
  with gr.Row():
@@ -365,6 +367,7 @@ with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base(primary_hu
365
  with gr.Row():
366
  mdx23c_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
367
  mdx23c_override_seg_size = gr.Checkbox(value=False, label="Override segment size", info="Override model default segment size instead of using the model default value.")
 
368
  mdx23c_overlap = gr.Slider(minimum=2, maximum=50, step=1, value=8, label="Overlap", info="Amount of overlap between prediction windows. Higher is better but slower.")
369
  mdx23c_pitch_shift = gr.Slider(minimum=-12, maximum=12, step=1, value=0, label="Pitch shift", info="Shift audio pitch by a number of semitones while processing. may improve output for deep/high vocals.")
370
  with gr.Row():
@@ -382,6 +385,7 @@ with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base(primary_hu
382
  with gr.Row():
383
  mdx_hop_length = gr.Slider(minimum=32, maximum=2048, step=32, value=1024, label="Hop Length", info="Usually called stride in neural networks; only change if you know what you're doing.")
384
  mdx_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
 
385
  mdx_overlap = gr.Slider(minimum=0.001, maximum=0.999, step=0.001, value=0.25, label="Overlap", info="Amount of overlap between prediction windows. Higher is better but slower.")
386
  mdx_denoise = gr.Checkbox(value=False, label="Denoise", info="Enable denoising after separation.")
387
  with gr.Row():
@@ -399,10 +403,12 @@ with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base(primary_hu
399
  with gr.Row():
400
  vr_window_size = gr.Slider(minimum=320, maximum=1024, step=32, value=512, label="Window Size", info="Balance quality and speed. 1024 = fast but lower, 320 = slower but better quality.")
401
  vr_aggression = gr.Slider(minimum=1, maximum=50, step=1, value=5, label="Agression", info="Intensity of primary stem extraction.")
402
- vr_tta = gr.Checkbox(value=False, label="TTA", info="Enable Test-Time-Augmentation; slow but improves quality.")
 
 
403
  vr_post_process = gr.Checkbox(value=False, label="Post Process", info="Identify leftover artifacts within vocal output; may improve separation for some songs.")
404
  vr_post_process_threshold = gr.Slider(minimum=0.1, maximum=0.3, step=0.1, value=0.2, label="Post Process Threshold", info="Threshold for post-processing.")
405
- vr_high_end_process = gr.Checkbox(value=False, label="High End Process", info="Mirror the missing frequency range of the output.")
406
  with gr.Row():
407
  vr_audio = gr.Audio(label="Input Audio", type="filepath")
408
  with gr.Row():
 
338
 
339
 
340
 
341
+ with gr.Blocks(title="🎵 Audio-Separator 🎵",theme=gr.themes.Base()) as app:
342
  gr.HTML("<h1> 🎵 Audio-Separator 🎵 </h1>")
343
 
344
  with gr.Tab("Roformer"):
 
346
  with gr.Row():
347
  roformer_model = gr.Dropdown(label="Select the Model", choices=list(ROFORMER_MODELS.keys()))
348
  with gr.Row():
349
+ with gr.Row():
350
+ roformer_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
351
+ roformer_override_seg_size = gr.Checkbox(value=False, label="Override segment size", info="Override model default segment size instead of using the model default value.")
352
+ with gr.Row():
353
+ roformer_overlap = gr.Slider(minimum=2, maximum=10, step=1, value=8, label="Overlap", info="Amount of overlap between prediction windows. Lower is better but slower.")
354
+ roformer_pitch_shift = gr.Slider(minimum=-12, maximum=12, step=1, value=0, label="Pitch shift", info="Shift audio pitch by a number of semitones while processing. may improve output for deep/high vocals.")
355
  with gr.Row():
356
  roformer_audio = gr.Audio(label="Input Audio", type="filepath")
357
  with gr.Row():
 
367
  with gr.Row():
368
  mdx23c_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
369
  mdx23c_override_seg_size = gr.Checkbox(value=False, label="Override segment size", info="Override model default segment size instead of using the model default value.")
370
+ with gr.Row():
371
  mdx23c_overlap = gr.Slider(minimum=2, maximum=50, step=1, value=8, label="Overlap", info="Amount of overlap between prediction windows. Higher is better but slower.")
372
  mdx23c_pitch_shift = gr.Slider(minimum=-12, maximum=12, step=1, value=0, label="Pitch shift", info="Shift audio pitch by a number of semitones while processing. may improve output for deep/high vocals.")
373
  with gr.Row():
 
385
  with gr.Row():
386
  mdx_hop_length = gr.Slider(minimum=32, maximum=2048, step=32, value=1024, label="Hop Length", info="Usually called stride in neural networks; only change if you know what you're doing.")
387
  mdx_seg_size = gr.Slider(minimum=32, maximum=4000, step=32, value=256, label="Segment Size", info="Larger consumes more resources, but may give better results.")
388
+ with gr.Row():
389
  mdx_overlap = gr.Slider(minimum=0.001, maximum=0.999, step=0.001, value=0.25, label="Overlap", info="Amount of overlap between prediction windows. Higher is better but slower.")
390
  mdx_denoise = gr.Checkbox(value=False, label="Denoise", info="Enable denoising after separation.")
391
  with gr.Row():
 
403
  with gr.Row():
404
  vr_window_size = gr.Slider(minimum=320, maximum=1024, step=32, value=512, label="Window Size", info="Balance quality and speed. 1024 = fast but lower, 320 = slower but better quality.")
405
  vr_aggression = gr.Slider(minimum=1, maximum=50, step=1, value=5, label="Agression", info="Intensity of primary stem extraction.")
406
+ with gr.Row():
407
+ vr_tta = gr.Checkbox(value=False, label="TTA", info="Enable Test-Time-Augmentation; slow but improves quality.")
408
+ with gr.Row():
409
  vr_post_process = gr.Checkbox(value=False, label="Post Process", info="Identify leftover artifacts within vocal output; may improve separation for some songs.")
410
  vr_post_process_threshold = gr.Slider(minimum=0.1, maximum=0.3, step=0.1, value=0.2, label="Post Process Threshold", info="Threshold for post-processing.")
411
+ vr_high_end_process = gr.Checkbox(value=False, label="High End Process", info="Mirror the missing frequency range of the output.")
412
  with gr.Row():
413
  vr_audio = gr.Audio(label="Input Audio", type="filepath")
414
  with gr.Row():