app.py

import gradio as gr
import moviepy.editor as mp
import numpy as np
import librosa
import matplotlib.pyplot as plt
from PIL import Image
import tempfile
import os

def generate_waveform_video(audio_file, image_file):
    try:
        # 1. Cargar audio
        y, sr = librosa.load(audio_file)
        duration = librosa.get_duration(y=y, sr=sr)

        # 2. Cargar imagen
        img_clip = mp.ImageClip(image_file).set_duration(duration)
        img_w, img_h = img_clip.size

        # 3. Crear efecto de waveform
        audio_envelope = np.abs(y)
        audio_envelope = (audio_envelope / np.max(audio_envelope)) * (img_h // 3)

        def make_frame(t):
            fig, ax = plt.subplots(figsize=(img_w/100, img_h/100), dpi=100)
            ax.set_xlim(0, duration)
            ax.set_ylim(-img_h//2, img_h//2)
            ax.axis('off')

            time_idx = int(t * sr)
            start = max(0, time_idx - sr//10)
            end = min(len(audio_envelope), time_idx + sr//10)
            wave_slice = audio_envelope[start:end]

            x = np.linspace(t-0.1, t+0.1, len(wave_slice))
            ax.fill_between(x, wave_slice - img_h//4, -wave_slice + img_h//4, 
                           facecolor='red', alpha=0.7)

            # Convertir figura a numpy array
            buf = io.BytesIO()
            plt.savefig(buf, format='png', bbox_inches='tight', pad_inches=0)
            plt.close(fig)
            return np.array(Image.open(buf))  # Formato correcto para MoviePy

        # Crear video
        effect_clip = mp.VideoClip(make_frame, duration=duration).set_fps(24)
        final_clip = mp.CompositeVideoClip([img_clip, effect_clip.set_pos("center")])
        final_clip = final_clip.set_audio(mp.AudioFileClip(audio_file))

        # Guardar en archivo temporal
        with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmpfile:
            final_clip.write_videofile(tmpfile.name, fps=24, codec="libx264", 
                                      audio_codec="aac", logger=None)
            return tmpfile.name

    except Exception as e:
        return f"Error: {str(e)}"

# Interfaz Gradio
iface = gr.Interface(
    fn=generate_waveform_video,
    inputs=[
        gr.Audio(type="filepath", label="Audio (WAV/MP3)"),
        gr.Image(type="filepath", label="Imagen de Fondo"),
    ],
    outputs=gr.Video(label="Video Resultante", format="mp4"),
    title="Generador de Video con Efectos de Audio",
    description="Crea videos con efectos visuales sincronizados con el audio. Actualmente soporta efecto de waveform."
)

if __name__ == "__main__":
    iface.queue().launch()