app.py

import gradio as gr
import torch
from diffusers import StableDiffusionPipeline
import imageio
from moviepy.editor import ImageSequenceClip

# Load the text-to-video model
def load_model():
    model = StableDiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-v1-4")
    model.to("cuda")  # Ensure the model runs on GPU
    return model

model = load_model()

# Generate video frames
def generate_video(prompt, num_frames=30, fps=10):
    frames = []
    for i in range(num_frames):
        # Add variation to the prompt for each frame
        frame_prompt = f"{prompt}, frame {i}"
        image = model(frame_prompt).images[0]
        frames.append(image)
    
    # Save frames as video
    video_path = "generated_video.mp4"
    clip = ImageSequenceClip([f for f in frames], fps=fps)
    clip.write_videofile(video_path, codec="libx264")
    
    return video_path

# Gradio Interface
def process_prompt(prompt):
    video_path = generate_video(prompt)
    return video_path

interface = gr.Interface(
    fn=process_prompt,
    inputs="text",
    outputs="video",
    title="Text-to-Video Generator",
    description="Enter a prompt to generate a video based on your description."
)

# Launch the app
interface.launch()