Spaces:
Running
on
Zero
Running
on
Zero
Peter Shi
commited on
Commit
Β·
42e2df6
1
Parent(s):
832604f
Revert to simple clean UI
Browse files
app.py
CHANGED
|
@@ -4,7 +4,6 @@ import torch
|
|
| 4 |
import torchaudio
|
| 5 |
import tempfile
|
| 6 |
import warnings
|
| 7 |
-
import os
|
| 8 |
warnings.filterwarnings("ignore")
|
| 9 |
|
| 10 |
from sam_audio import SAMAudio, SAMAudioProcessor
|
|
@@ -49,150 +48,48 @@ def separate_audio(audio_path, video_path, text_prompt):
|
|
| 49 |
target_path = save_audio(result.target[0].unsqueeze(0).cpu(), sample_rate)
|
| 50 |
residual_path = save_audio(result.residual[0].unsqueeze(0).cpu(), sample_rate)
|
| 51 |
|
| 52 |
-
return target_path, residual_path, f"β
Successfully isolated
|
| 53 |
|
| 54 |
except Exception as e:
|
| 55 |
import traceback
|
| 56 |
traceback.print_exc()
|
| 57 |
return None, None, f"β Error: {str(e)}"
|
| 58 |
|
| 59 |
-
#
|
| 60 |
-
|
| 61 |
-
.
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
}
|
| 71 |
-
|
| 72 |
-
input, textarea {
|
| 73 |
-
background: #1a1a1a !important;
|
| 74 |
-
border: 1px solid #444 !important;
|
| 75 |
-
color: #ffffff !important;
|
| 76 |
-
}
|
| 77 |
-
|
| 78 |
-
input::placeholder, textarea::placeholder {
|
| 79 |
-
color: #888 !important;
|
| 80 |
-
}
|
| 81 |
-
|
| 82 |
-
.primary-btn {
|
| 83 |
-
background: linear-gradient(135deg, #e91e8c, #9c27b0) !important;
|
| 84 |
-
border: none !important;
|
| 85 |
-
border-radius: 24px !important;
|
| 86 |
-
color: #ffffff !important;
|
| 87 |
-
font-weight: 600 !important;
|
| 88 |
-
}
|
| 89 |
-
|
| 90 |
-
.example-btn {
|
| 91 |
-
background: #2a2a2a !important;
|
| 92 |
-
border: 1px solid #444 !important;
|
| 93 |
-
color: #ffffff !important;
|
| 94 |
-
border-radius: 8px !important;
|
| 95 |
-
}
|
| 96 |
-
|
| 97 |
-
.example-btn:hover {
|
| 98 |
-
background: #3a3a3a !important;
|
| 99 |
-
border-color: #e91e8c !important;
|
| 100 |
-
}
|
| 101 |
-
|
| 102 |
-
hr {
|
| 103 |
-
border-color: #333 !important;
|
| 104 |
-
}
|
| 105 |
-
"""
|
| 106 |
-
|
| 107 |
-
# Build Gradio Interface
|
| 108 |
-
with gr.Blocks(
|
| 109 |
-
title="SAM-Audio - Isolate Sounds",
|
| 110 |
-
theme=gr.themes.Default(
|
| 111 |
-
primary_hue="pink",
|
| 112 |
-
secondary_hue="purple",
|
| 113 |
-
neutral_hue="gray",
|
| 114 |
-
),
|
| 115 |
-
css=custom_css
|
| 116 |
-
) as demo:
|
| 117 |
|
| 118 |
with gr.Row():
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
gr.
|
| 122 |
-
gr.
|
| 123 |
-
|
| 124 |
-
gr.Markdown("---")
|
| 125 |
-
gr.Markdown("### How it works")
|
| 126 |
-
gr.Markdown("**1.** Add audio or video")
|
| 127 |
-
gr.Markdown("**2.** Describe the sound")
|
| 128 |
-
gr.Markdown("**3.** Get separated tracks")
|
| 129 |
|
| 130 |
-
gr.Markdown("---")
|
| 131 |
-
gr.Markdown("**Model:** SAM-Audio Small")
|
| 132 |
-
|
| 133 |
-
# Main content area
|
| 134 |
-
with gr.Column(scale=4):
|
| 135 |
-
gr.Markdown("## π€ Upload Audio or Video")
|
| 136 |
-
|
| 137 |
-
with gr.Tabs():
|
| 138 |
-
with gr.TabItem("οΏ½ Audio"):
|
| 139 |
-
input_audio = gr.Audio(
|
| 140 |
-
label="Upload audio file (MP3, WAV, FLAC, etc.)",
|
| 141 |
-
type="filepath"
|
| 142 |
-
)
|
| 143 |
-
|
| 144 |
-
with gr.TabItem("π¬ Video"):
|
| 145 |
-
input_video = gr.Video(
|
| 146 |
-
label="Upload video file (MP4, MKV, AVI, etc.)"
|
| 147 |
-
)
|
| 148 |
-
|
| 149 |
-
gr.Markdown("## π¬ Describe the Sound")
|
| 150 |
text_prompt = gr.Textbox(
|
| 151 |
-
label="",
|
| 152 |
placeholder="e.g., 'A man speaking', 'Piano melody', 'Dog barking'",
|
| 153 |
-
|
| 154 |
-
)
|
| 155 |
-
|
| 156 |
-
run_btn = gr.Button(
|
| 157 |
-
"π― Isolate Sound",
|
| 158 |
-
variant="primary",
|
| 159 |
-
size="lg",
|
| 160 |
-
elem_classes=["primary-btn"]
|
| 161 |
-
)
|
| 162 |
-
|
| 163 |
-
status_output = gr.Markdown(
|
| 164 |
-
value="Upload a file and describe what sound you want to isolate."
|
| 165 |
)
|
| 166 |
|
| 167 |
-
gr.
|
| 168 |
-
gr.Markdown("
|
| 169 |
-
|
| 170 |
-
|
| 171 |
-
|
| 172 |
-
|
| 173 |
-
|
| 174 |
-
|
| 175 |
-
|
| 176 |
-
|
| 177 |
-
|
| 178 |
-
|
| 179 |
-
gr.Markdown("---")
|
| 180 |
-
gr.Markdown("## π‘ Example Prompts")
|
| 181 |
-
|
| 182 |
-
with gr.Row():
|
| 183 |
-
btn1 = gr.Button("A man speaking", elem_classes=["example-btn"])
|
| 184 |
-
btn2 = gr.Button("A woman singing", elem_classes=["example-btn"])
|
| 185 |
-
btn3 = gr.Button("Piano", elem_classes=["example-btn"])
|
| 186 |
-
btn4 = gr.Button("Drums", elem_classes=["example-btn"])
|
| 187 |
-
btn5 = gr.Button("Guitar", elem_classes=["example-btn"])
|
| 188 |
-
btn6 = gr.Button("Dog barking", elem_classes=["example-btn"])
|
| 189 |
-
|
| 190 |
-
btn1.click(fn=lambda: "A man speaking", outputs=[text_prompt])
|
| 191 |
-
btn2.click(fn=lambda: "A woman singing", outputs=[text_prompt])
|
| 192 |
-
btn3.click(fn=lambda: "Piano", outputs=[text_prompt])
|
| 193 |
-
btn4.click(fn=lambda: "Drums", outputs=[text_prompt])
|
| 194 |
-
btn5.click(fn=lambda: "Guitar", outputs=[text_prompt])
|
| 195 |
-
btn6.click(fn=lambda: "Dog barking", outputs=[text_prompt])
|
| 196 |
|
| 197 |
run_btn.click(
|
| 198 |
fn=separate_audio,
|
|
|
|
| 4 |
import torchaudio
|
| 5 |
import tempfile
|
| 6 |
import warnings
|
|
|
|
| 7 |
warnings.filterwarnings("ignore")
|
| 8 |
|
| 9 |
from sam_audio import SAMAudio, SAMAudioProcessor
|
|
|
|
| 48 |
target_path = save_audio(result.target[0].unsqueeze(0).cpu(), sample_rate)
|
| 49 |
residual_path = save_audio(result.residual[0].unsqueeze(0).cpu(), sample_rate)
|
| 50 |
|
| 51 |
+
return target_path, residual_path, f"β
Successfully isolated '{text_prompt}'"
|
| 52 |
|
| 53 |
except Exception as e:
|
| 54 |
import traceback
|
| 55 |
traceback.print_exc()
|
| 56 |
return None, None, f"β Error: {str(e)}"
|
| 57 |
|
| 58 |
+
# Build Gradio Interface - Simple and clean
|
| 59 |
+
with gr.Blocks(title="SAM-Audio Demo") as demo:
|
| 60 |
+
gr.Markdown(
|
| 61 |
+
"""
|
| 62 |
+
# π΅ SAM-Audio: Segment Anything for Audio
|
| 63 |
+
|
| 64 |
+
Isolate specific sounds from an audio or video file using natural language prompts.
|
| 65 |
+
|
| 66 |
+
**Model:** [facebook/sam-audio-small](https://huggingface.co/facebook/sam-audio-small)
|
| 67 |
+
"""
|
| 68 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
with gr.Row():
|
| 71 |
+
with gr.Column():
|
| 72 |
+
gr.Markdown("### Upload Audio or Video")
|
| 73 |
+
input_audio = gr.Audio(label="Audio File", type="filepath")
|
| 74 |
+
input_video = gr.Video(label="Video File (MP4)")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 76 |
text_prompt = gr.Textbox(
|
| 77 |
+
label="Text Prompt",
|
| 78 |
placeholder="e.g., 'A man speaking', 'Piano melody', 'Dog barking'",
|
| 79 |
+
info="Describe the sound you want to isolate."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
)
|
| 81 |
|
| 82 |
+
run_btn = gr.Button("π― Isolate Sound", variant="primary")
|
| 83 |
+
status_output = gr.Markdown(value="")
|
| 84 |
+
|
| 85 |
+
with gr.Column():
|
| 86 |
+
gr.Markdown("### Results")
|
| 87 |
+
output_target = gr.Audio(label="Isolated Sound (Target)")
|
| 88 |
+
output_residual = gr.Audio(label="Background (Residual)")
|
| 89 |
+
|
| 90 |
+
gr.Markdown("---")
|
| 91 |
+
gr.Markdown("### Example Prompts")
|
| 92 |
+
gr.Markdown("- A man speaking\n- A woman singing\n- Piano\n- Drums\n- Guitar\n- Dog barking\n- Car engine")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
|
| 94 |
run_btn.click(
|
| 95 |
fn=separate_audio,
|