Update app.py

app.py

@@ -1,44 +1,85 @@
 import gradio as gr
 from pydub import AudioSegment
-import noisereduce as nr
 import numpy as np
 import tempfile
 import os
+import noisereduce as nr
+from scipy.io import wavfile
+import subprocess
+import shutil
+
+# Helper functions
+def audiosegment_to_array(audio):
+    return np.array(audio.get_array_of_samples()), audio.frame_rate
 
-# Helper: Convert AudioSegment to numpy array and back
-def audiosegment_to_array(audio_segment):
-    return np.array(audio_segment.get_array_of_samples()), audio_segment.frame_rate
+def array_to_audiosegment(samples, frame_rate, sample_width=2, channels=1):
+    return AudioSegment(
+        samples.tobytes(),
+        frame_rate=frame_rate,
+        sample_width=sample_width,
+        channels=channels
+    )
 
-# Apply effect functions
+# Effect Functions
 def apply_normalize(audio):
     return audio.normalize()
 
 def apply_noise_reduction(audio):
     samples, frame_rate = audiosegment_to_array(audio)
-    reduced_noise = nr.reduce_noise(y=samples, sr=frame_rate)
-    return AudioSegment(
-        reduced_noise.tobytes(),
-        frame_rate=frame_rate,
-        sample_width=reduced_noise.dtype.itemsize,
-        channels=audio.channels
-    )
+    reduced = nr.reduce_noise(y=samples, sr=frame_rate)
+    return array_to_audiosegment(reduced, frame_rate, channels=audio.channels)
 
 def apply_compression(audio):
     return audio.compress_dynamic_range()
 
 def apply_reverb(audio):
-    # Simulate reverb by overlaying a delayed, quieter copy
-    reverb_audio = audio - 10  # Lower volume
-    return audio.overlay(reverb_audio, position=1000)
+    reverb = audio - 10
+    return audio.overlay(reverb, position=1000)
 
 def apply_pitch_shift(audio, semitones=-2):
-    return audio._spawn(audio._array_type(
-        np.interp(
-            np.arange(0, len(audio), 2 ** (semitones / 12)),
-            np.arange(len(audio)),
-            audio.get_array_of_samples()
-        ).astype(np.int16).tolist()
-    ), overrides={"frame_rate": int(audio.frame_rate * (2 ** (semitones / 12)))})
+    new_frame_rate = int(audio.frame_rate * (2 ** (semitones / 12)))
+    samples = np.array(audio.get_array_of_samples())
+    resampled = np.interp(
+        np.arange(0, len(samples), 2 ** (semitones / 12)),
+        np.arange(len(samples)),
+        samples
+    ).astype(np.int16)
+    return AudioSegment(
+        resampled.tobytes(),
+        frame_rate=new_frame_rate,
+        sample_width=audio.sample_width,
+        channels=audio.channels
+    )
+
+def apply_echo(audio, delay_ms=500, decay=0.5):
+    echo = audio - 10
+    return audio.overlay(echo, position=delay_ms)
+
+def apply_stereo_widen(audio, pan_amount=0.3):
+    left = audio.pan(-pan_amount)
+    right = audio.pan(pan_amount)
+    return AudioSegment.from_mono_audiosegments(left, right)
+
+def apply_bass_boost(audio, gain=10):
+    return audio.low_pass_filter(100).apply_gain(gain)
+
+def apply_treble_boost(audio, gain=10):
+    return audio.high_pass_filter(4000).apply_gain(gain)
+
+# Save temp WAV file for spleeter
+def save_temp_wav(audio, path):
+    audio.export(path, format="wav")
+
+# Vocal Isolation using spleeter
+def apply_vocal_isolation(audio_path):
+    output_dir = tempfile.mkdtemp()
+    cmd = ["spleeter", "separate", "-o", output_dir, audio_path]
+    subprocess.run(cmd)
+    vocal_path = os.path.join(output_dir, os.path.basename(audio_path).replace(".wav", ""), "vocals.wav")
+    if os.path.exists(vocal_path):
+        return vocal_path
+    else:
+        raise Exception("Vocal isolation failed.")
 
 # Main processing function
 def process_audio(audio_file, effect):
@@ -53,9 +94,22 @@ def process_audio(audio_file, effect):
     elif effect == "Add Reverb":
         result = apply_reverb(audio)
     elif effect == "Pitch Shift":
-        result = apply_pitch_shift(audio, semitones=-2)  # Down 2 semitones
+        result = apply_pitch_shift(audio)
+    elif effect == "Echo":
+        result = apply_echo(audio)
+    elif effect == "Stereo Widening":
+        result = apply_stereo_widen(audio)
+    elif effect == "Bass Boost":
+        result = apply_bass_boost(audio)
+    elif effect == "Treble Boost":
+        result = apply_treble_boost(audio)
+    elif effect == "Isolate Vocals":
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
+            save_temp_wav(audio, f.name)
+            vocal_path = apply_vocal_isolation(f.name)
+            result = AudioSegment.from_wav(vocal_path)
     else:
-        result = audio  # No effect
+        result = audio
 
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
         result.export(f.name, format="wav")
@@ -72,14 +126,19 @@ interface = gr.Interface(
                 "Noise Reduction",
                 "Compress Dynamic Range",
                 "Add Reverb",
-                "Pitch Shift"
+                "Pitch Shift",
+                "Echo",
+                "Stereo Widening",
+                "Bass Boost",
+                "Treble Boost",
+                "Isolate Vocals"
             ],
             label="Select Effect"
         )
     ],
     outputs=gr.Audio(label="Processed Audio", type="filepath"),
-    title="Fix My Recording - Advanced Edition",
-    description="Upload your audio to clean, master, and add creative effects with AI!"
+    title="Fix My Recording - Pro Edition",
+    description="Apply noise reduction, mastering, creative effects, or isolate vocals from your track!"
 )
 
 interface.launch()