app.py

import gradio as gr
import numpy as np
from PIL import Image
import cv2
from moviepy.editor import VideoFileClip
from share_btn import community_icon_html, loading_icon_html, share_js
import torch
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
from diffusers.utils import export_to_video

pipe_xl = DiffusionPipeline.from_pretrained("cerspense/zeroscope_v2_XL", torch_dtype=torch.float16, revision="refs/pr/17")
pipe_xl.vae.enable_slicing()
pipe_xl.scheduler = DPMSolverMultistepScheduler.from_config(pipe_xl.scheduler.config)
pipe_xl.enable_model_cpu_offload()
pipe_xl.to("cuda")

def convert_mp4_to_frames(video_path, duration=3):
    # Read the video file
    video = cv2.VideoCapture(video_path)

    # Get the frames per second (fps) of the video
    fps = video.get(cv2.CAP_PROP_FPS)

    # Calculate the number of frames to extract
    num_frames = int(fps * duration)

    frames = []
    frame_count = 0
    
    # Iterate through each frame
    while True:
        # Read a frame
        ret, frame = video.read()
        
        # If the frame was not successfully read or we have reached the desired duration, break the loop
        if not ret or frame_count == num_frames:
            break
        
        # Convert BGR to RGB
        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

        # Append the frame to the list of frames
        frames.append(frame)

        frame_count += 1

    # Release the video object
    video.release()

    # Convert the list of frames to a numpy array
    frames = np.array(frames)

    return frames

def infer(prompt, video_in, denoise_strength):

    negative_prompt = "text, watermark, copyright, blurry, nsfw"

    video = convert_mp4_to_frames(video_in, duration=3)
    video_resized = [Image.fromarray(frame).resize((1024, 576)) for frame in video]
    video_frames = pipe_xl(prompt, negative_prompt=negative_prompt, video=video_resized, strength=denoise_strength).frames
    video_path = export_to_video(video_frames, output_video_path="xl_result.mp4")
    
    return "xl_result.mp4", gr.Group.update(visible=True)

css = """
#col-container {max-width: 510px; margin-left: auto; margin-right: auto;}
a {text-decoration-line: underline; font-weight: 600;}
.animate-spin {
  animation: spin 1s linear infinite;
}

@keyframes spin {
  from {
      transform: rotate(0deg);
  }
  to {
      transform: rotate(360deg);
  }
}

#share-btn-container {
  display: flex; 
  padding-left: 0.5rem !important; 
  padding-right: 0.5rem !important; 
  background-color: #000000; 
  justify-content: center; 
  align-items: center; 
  border-radius: 9999px !important; 
  max-width: 13rem;
}

#share-btn-container:hover {
  background-color: #060606;
}

#share-btn {
  all: initial; 
  color: #ffffff;
  font-weight: 600; 
  cursor:pointer; 
  font-family: 'IBM Plex Sans', sans-serif; 
  margin-left: 0.5rem !important; 
  padding-top: 0.5rem !important; 
  padding-bottom: 0.5rem !important;
  right:0;
}

#share-btn * {
  all: unset;
}

#share-btn-container div:nth-child(-n+2){
  width: auto !important;
  min-height: 0px !important;
}

#share-btn-container .wrap {
  display: none !important;
}

#share-btn-container.hidden {
  display: none!important;
}
img[src*='#center'] { 
    display: block;
    margin: auto;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(
            """
            <h1 style="text-align: center;">Zeroscope XL</h1>
            <p style="text-align: center;">
            This space is specifically designed for upscaling content made from <br />
            <a href="https://huggingface.co/spaces/fffiloni/zeroscope">the zeroscope_v2_576w space</a> using vid2vid. <br />
            Remember to use the same prompt that was used to generate the original clip.<br />
            For demo purpose, video length is limited to 3 seconds.
            </p>
            
            [![Duplicate this Space](https://huggingface.co/datasets/huggingface/badges/raw/main/duplicate-this-space-sm.svg#center)](https://huggingface.co/spaces/fffiloni/zeroscope-XL?duplicate=true)
            
            """
        )

        video_in = gr.Video(type="numpy", source="upload")
        prompt_in = gr.Textbox(label="Prompt", placeholder="This must be the same prompt you used for the original clip :)", elem_id="prompt-in")
        denoise_strength = gr.Slider(label="Denoise strength", minimum=0.6, maximum=0.9, step=0.01, value=0.66)
        #inference_steps = gr.Slider(label="Inference Steps", minimum=10, maximum=100, step=1, value=40, interactive=False)
        submit_btn = gr.Button("Submit")
        video_result = gr.Video(label="Video Output", elem_id="video-output")

        with gr.Group(elem_id="share-btn-container", visible=False) as share_group:
            community_icon = gr.HTML(community_icon_html)
            loading_icon = gr.HTML(loading_icon_html)
            share_button = gr.Button("Share to community", elem_id="share-btn")

    submit_btn.click(fn=infer,
                    inputs=[prompt_in, video_in, denoise_strength],
                    outputs=[video_result, share_group])
    
    share_button.click(None, [], [], _js=share_js)

demo.queue(max_size=12).launch()