Update app.py

app.py

@@ -4,16 +4,14 @@ import numpy as np
 import tempfile
 import os
 import noisereduce as nr
-from scipy.io import wavfile
-import subprocess
+import json
 import torch
 from demucs import pretrained
 from demucs.apply import apply_model
 import torchaudio
-import torch
 from pathlib import Path
 
-# Helper functions
+# === Helper Functions ===
 def audiosegment_to_array(audio):
     return np.array(audio.get_array_of_samples()), audio.frame_rate
 
@@ -25,7 +23,7 @@ def array_to_audiosegment(samples, frame_rate, channels=1):
         channels=channels
     )
 
-# Effect Functions
+# === Effect Functions ===
 def apply_normalize(audio):
     return audio.normalize()
 
@@ -71,33 +69,7 @@ def apply_bass_boost(audio, gain=10):
 def apply_treble_boost(audio, gain=10):
     return audio.high_pass_filter(4000).apply_gain(gain)
 
-# Vocal Isolation using Demucs
-def apply_vocal_isolation(audio_path):
-    model = pretrained.get_model(name='htdemucs')
-    wav = load_track_local(audio_path, model.samplerate, channels=2)  # stereo
-    ref = wav.mean(0)
-    wav -= ref[:, None]
-    sources = apply_model(model, wav[None])[0]
-    wav += ref[:, None]
-
-    # Get vocals (index 3)
-    vocal_track = sources[3].cpu()
-
-    out_path = os.path.join(tempfile.gettempdir(), "vocals.wav")
-    save_track(out_path, vocal_track, model.samplerate)
-    return out_path
-
-
-# Local copy of helper functions from demucs
-def load_track(track, sample_rate, mono=True):
-    wav, sr = torchaudio.load(str(track))
-    if sr != sample_rate:
-        wav = torchaudio.functional.resample(wav, sr, sample_rate)
-    if mono and wav.shape[0] == 2:
-        wav = wav.mean(0)
-    return wav
-
-
+# === Vocal Isolation Helpers ===
 def load_track_local(path, sample_rate, channels=2):
     sig, rate = torchaudio.load(path)
     if rate != sample_rate:
@@ -106,16 +78,38 @@ def load_track_local(path, sample_rate, channels=2):
         sig = sig.mean(0)
     return sig
 
-
 def save_track(path, wav, sample_rate):
     path = Path(path)
     torchaudio.save(str(path), wav, sample_rate)
+
+def apply_vocal_isolation(audio_path):
+    model = pretrained.get_model(name='htdemucs')
+    wav = load_track_local(audio_path, model.samplerate, channels=2)
+    ref = wav.mean(0)
+    wav -= ref[:, None]
+    sources = apply_model(model, wav[None])[0]
+    wav += ref[:, None]
+
+    vocal_track = sources[3].cpu()  # index 3 = vocals
+    out_path = os.path.join(tempfile.gettempdir(), "vocals.wav")
+    save_track(out_path, vocal_track, model.samplerate)
     return out_path
 
-# Apply selected effects in order
-def process_audio(audio_file, effects, isolate_vocals):
+# === Preset Loader ===
+def load_presets():
+    preset_files = [f for f in os.listdir("presets") if f.endswith(".json")]
+    presets = {}
+    for f in preset_files:
+        with open(os.path.join("presets", f)) as infile:
+            data = json.load(infile)
+            presets[data["name"]] = data["effects"]
+    return presets
+
+preset_choices = load_presets()
+
+# === Main Processing Function ===
+def process_audio(audio_file, selected_effects, isolate_vocals, preset_name):
     audio = AudioSegment.from_file(audio_file)
-    original = audio
 
     effect_map = {
         "Noise Reduction": apply_noise_reduction,
@@ -129,7 +123,9 @@ def process_audio(audio_file, effects, isolate_vocals):
         "Normalize": apply_normalize,
     }
 
-    for effect_name in effects:
+    # Apply selected preset or custom effects
+    effects_to_apply = preset_choices.get(preset_name, selected_effects)
+    for effect_name in effects_to_apply:
         if effect_name in effect_map:
             audio = effect_map[effect_name](audio)
 
@@ -142,11 +138,12 @@ def process_audio(audio_file, effects, isolate_vocals):
         else:
             final_audio = audio
 
-        final_audio.export(f.name, format="wav")
-        return f.name
+        output_path = f.name
+        final_audio.export(output_path, format="mp3")
+        return output_path
 
-# Gradio Interface
-effect_choices = [
+# === Gradio Interface ===
+effect_options = [
     "Noise Reduction",
     "Compress Dynamic Range",
     "Add Reverb",
@@ -158,16 +155,19 @@ effect_choices = [
     "Normalize"
 ]
 
+preset_names = list(preset_choices.keys())
+
 interface = gr.Interface(
     fn=process_audio,
     inputs=[
         gr.Audio(label="Upload Audio", type="filepath"),
-        gr.CheckboxGroup(choices=effect_choices, label="Apply Effects in Order"),
-        gr.Checkbox(label="Isolate Vocals After Effects")
+        gr.CheckboxGroup(choices=effect_options, label="Apply Effects in Order"),
+        gr.Checkbox(label="Isolate Vocals After Effects"),
+        gr.Dropdown(choices=preset_names, label="Select Preset", value=preset_names[0] if preset_names else None)
     ],
-    outputs=gr.Audio(label="Processed Audio", type="filepath"),
-    title="Fix My Recording - Studio Pro",
-    description="Apply multiple effects in sequence and optionally isolate vocals!",
+    outputs=gr.Audio(label="Processed Audio (MP3)", type="filepath"),
+    title="AI Audio Studio - Pro Edition",
+    description="Apply multiple effects, isolate vocals, and export polished tracks -- all powered by AI!",
     allow_flagging="never"
 )