utils.py

import numpy as np
from PIL import Image
import tempfile
from moviepy.editor import ImageClip, AudioArrayClip
import scipy

def create_video(image: "PIL.Image", audio: tuple, duration: float) -> str:
    # Convert image to numpy array
    img_array = np.array(image)
    
    # Unpack audio data
    sample_rate, audio_data = audio
    
    # Ensure audio is the correct duration
    target_samples = int(duration * sample_rate)
    if audio_data.shape[1] > target_samples:
        audio_data = audio_data[:, :target_samples]
    else:
        # Pad audio if too short
        padding = target_samples - audio_data.shape[1]
        audio_data = np.pad(audio_data, ((0, 0), (0, padding)), mode='constant')
    
    # Create video clip
    img_clip = ImageClip(img_array, duration=duration)
    
    # Create audio clip
    audio_clip = AudioArrayClip(audio_data.T, fps=sample_rate)
    img_clip = img_clip.set_audio(audio_clip)
    
    # Save to temporary file
    with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmpfile:
        img_clip.write_videofile(
            tmpfile.name,
            fps=24,
            audio_codec="aac",
            logger=None
        )
        return tmpfile.name