remove axis + refactor total frames

app.py

@@ -2,52 +2,57 @@ import gradio as gr
 import numpy as np
 import matplotlib.pyplot as plt
 from matplotlib.animation import FuncAnimation
+import io
 import librosa
 import tempfile
 
-def extract_waveform_animation(audio_file, window_seconds=5, bg_color='black'):
+def extract_waveform_animation(audio_file, window_seconds=5):
     y, sr = librosa.load(audio_file, sr=None)
     duration = librosa.get_duration(y=y, sr=sr)
-    window_length = int(window_seconds * sr)
     
     fig, ax = plt.subplots()
-    line, = ax.plot([], [], lw=2, color='white')
+    line, = ax.plot([], [], lw=2)
+    window_length = int(window_seconds * sr)
     
-    # Remove all axes
+    # Initialize with first window
+    first_window = y[:window_length]
+    x_vals = np.linspace(0, duration, num=len(y))
     ax.set_axis_off()
-    ax.set_facecolor(bg_color)
     
     def init():
         ax.set_xlim(0, window_seconds)
+        ax.set_ylim(np.min(y), np.max(y))  # Reduced max for visibility
         return line,
 
     def update(frame):
+        # Get current window
         start = frame * sr
         end = start + window_length
         window = y[start:end]
-        x_vals = np.linspace(frame, frame + window_seconds, num=len(window))
         
-        line.set_data(x_vals, window)
+        # Update x and y limits
+        ax.set_xlim(frame, frame + window_seconds)
+        
+        # Update line data
+        line.set_data(x_vals[start:end], window)
         return line,
 
-    total_frames = int(duration) - window_seconds + 1
-    ani = FuncAnimation(fig, update, frames=np.arange(total_frames), 
-                       init_func=init, blit=False)
+    total_frames = int(duration)
+    ani = FuncAnimation(fig, update, frames=range(total_frames), 
+                       init_func=init, interval=7, blit=False)
     
     with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as tmpfile:
-        ani.save(tmpfile.name, writer='ffmpeg', fps=1, dpi=200)
+        ani.save(tmpfile.name, writer='ffmpeg', fps=1)
         video_path = tmpfile.name
     
-    plt.close(fig)
     return video_path
 
-# Modified interface with updated parameters
+# Modified interface with window controls
 iface = gr.Interface(
     fn=extract_waveform_animation,
     inputs=[
         gr.Audio(type="filepath"),
-        gr.Slider(1, 10, value=2, step=1, label="Window Size (seconds)"),
-        gr.ColorPicker(label="Background Color")
+        gr.Slider(1, 10, value=5, step=1, label="Window Size (seconds)")
     ],
     outputs=gr.Video(),
     description="Scroll through audio waveform with a moving window."