How to Create High-Quality Videos with Zeroscope_v2 and Zeroscope_v2_XL

Are you looking to create stunning, high-definition videos using deep learning models? If so, you’re in the right place. In this tutorial, I’ll guide you through using zeroscope_v2 576w and zeroscope_v2 XL to produce and upscale videos. These models are perfect for creating 16:9 compositions with smooth video output.

What is Zeroscope_v2 576w?

Zeroscope_v2 576w is a watermark-free Modelscope-based video model. It’s optimized for producing 576×320 resolution videos at 24 frames. The model has been trained on over 9,000 clips, making it adept at handling a variety of scenarios.

Zeroscope_v2 XL: The Upscaler

For those seeking even higher resolution, zeroscope_v2 XL is your go-to. It’s designed to upscale content from zeroscope_v2 576w to a crisp 1024×576 resolution, providing an even more detailed and refined output.

Using the Models

Generating a Video with zeroscope_v2 576w

  1. Install Necessary Libraries:
$ pip install diffusers transformers accelerate torch

2. Generate Your Video:

import torch
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
from diffusers.utils import export_to_video

pipe = DiffusionPipeline.from_pretrained("cerspense/zeroscope_v2_576w", torch_dtype=torch.float16)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)

prompt = "Your creative prompt here"
video_frames = pipe(prompt, num_inference_steps=40, height=320, width=576, num_frames=24).frames
video_path = export_to_video(video_frames)

Upscaling with zeroscope_v2 XL

  1. Generate a Low-Resolution Video: Follow the previous steps to create a video with zeroscope_v2 576w.
  2. Upscale Your Video:
pipe = DiffusionPipeline.from_pretrained("cerspense/zeroscope_v2_XL", torch_dtype=torch.float16)
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)

video = [Image.fromarray(frame).resize((1024, 576)) for frame in video_frames]
video_frames = pipe(prompt, video=video, strength=0.6).frames
video_path = export_to_video(video_frames, output_video_path="path_to_save_video")

Tips for Best Results

  • VRAM Usage: zeroscope_v2 576w uses 7.9GB of VRAM, while zeroscope_v2 XL requires 15.3GB.
  • Denoise Strength: Aim for a denoise strength between 0.66 and 0.85 when upscaling.
  • Frame Rate and Resolution: Stick to 24 frames and recommended resolutions for optimal results.


Using zeroscope_v2 576w and zeroscope_v2 XL, you can create and upscale videos with impressive quality. The process might seem complex at first, but once you get the hang of it, you’ll be producing professional-level videos in no time.


Leave a Reply

Your email address will not be published. Required fields are marked *