import gradio as gr
import spaces
import torch
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
from diffusers.utils import export_to_video
import cv2
import numpy as np

pipe = DiffusionPipeline.from_pretrained("damo-vilab/text-to-video-ms-1.7b", torch_dtype=torch.float16, variant="fp16")
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
pipe.enable_model_cpu_offload()
pipe.enable_vae_slicing()

@spaces.GPU(duration=250)
def generate(prompt, num_inference_steps, num_frames):
    video_frames = pipe(prompt, num_inference_steps, num_frames).frames
    video_path = export_to_video(video_frames)
    return video_path

prompt = gr.Textbox(label="Enter prompt to generate a video")
num_inference_steps = gr.Slider(10, 50, value=25)
num_frames = gr.Slider(100, 500, value=200)

interface = gr.Interface(
  generate,
  inputs=[prompt, num_inference_steps, num_frames],
  examples=[["Astronaut riding a horse", 24, 256], ["Darth vader surfing in waves", 32, 128]],
  outputs="video",
  cache_examples=False,
  theme="soft"
).launch()