app.py

import gradio as gr
import os
import librosa
import soundfile as sf
import subprocess
import uuid

OUTPUT_DIR = "segments"
os.makedirs(OUTPUT_DIR, exist_ok=True)

def process_vocal(file):
    filename = file.name
    y, sr = librosa.load(filename, sr=32000, mono=True)
    intervals = librosa.effects.split(y, top_db=25)

    segment_paths = []
    for i, (start, end) in enumerate(intervals):
        clip = y[start:end]
        if len(clip) / sr < 0.5 or len(clip) / sr > 6.0:
            continue
        out_path = os.path.join(OUTPUT_DIR, f"clip_{i}.wav")
        sf.write(out_path, clip, sr)
        segment_paths.append(out_path)

    zipname = f"vocals_{uuid.uuid4().hex[:8]}.zip"
    subprocess.run(["zip", "-j", zipname] + segment_paths)
    return zipname

gr.Interface(
    fn=process_vocal,
    inputs=gr.Audio(type="file"),
    outputs=gr.File(label="Download ZIP"),
    title="VocalPrep AI",
    description="Upload a vocal-only WAV file. This tool removes silence, splits usable clips, and gives you a training-ready ZIP."
).launch()