
Introduction
Show-1 is a deep learning model designed to generate realistic videos from text prompts. The model uses both latent-based and pixel-based text-to-video diffusion models (VDMs) to generate videos. While creating videos, the Show-1 model uses the pixel-based VDM to generate a low-resolution video that accurately matches the text prompt and applies the latent-based VDM to increase the video resolution.
This guide explains how to generate AI videos using the Show-1 model on a Vultr Cloud GPU server. You will set up the server environment with the necessary transformers, generative adversarial networks, and optical flow estimation libraries to generate high-quality videos that match your input text prompts.
Prerequisites
Before you begin:
Deploy a new Ubuntu 22.04 A100 Vultr Cloud GPU server with at least
80 GB of GPU RAM.Access the server using SSH as a non-root user with sudo privileges.
Connect and allow your Hugging Face account to use the DeepFloyd model.
Set Up the Server
Follow the steps below to install all required Show-1 dependency packages such as Gradio to process multimedia content on the server using the model.
Clone the Show-1 generative-model repository using Git.
console
List your directory files and verify that a new
Show-1 directory is available. console
Output:
Switch to the
Show-1 project directory. console
Install the Python virtual environment package using APT.
console
Create a new Python virtual environment such as
show1-env. console
Activate the virtual environment.
console
Update the Pip package manager.
console
Install Gradio using Pip.
console
Install the OpenCV library to export generated videos.
console
Back up the original requirements file.
console
Open the requirements file using a text editor such as Nano.
console
Add the following contents to the file.
The above configuration removes the version restrictions for
accelerate, torch, and torchvision modules. The original package versions may not be compatible with each other and removing the version specifications allows Pip to solve dependency conflicts.Install all required Python packages using the
requirements.txt file. console
Enable Access to the DeepFloyd Super-Resolution Model
The Show-1 AI video generator uses pixel-based VDMs that include a DeepFloyd super-resolution model to transform the initial low-resolution to a high-resolution video. Log in to your Hugging Face account using your access token on the server to enable access to the DeepFloyd model repository.
Log in to your Hugging Face account using the
huggingface-cli Transformers library. console
Enter your Hugging Face access token when prompted and press Enter to save changes.
When prompted to add the token as a Git credential, enter N and press Enter to apply changes. When successful, your output should look like the one below.
Back up the original model application file
app.py. console
Create a new Python application file
app.py. console
Add the following code to the file.
python
Save and close the file.
Below is what the above application functions do:
os.environ["HF_HOME"]: Sets up an environment variable with the Hugging Face home directory where your access token is stored.TextToVideoIFPipeline.from_pretrained(): Initializes a pretrained text-to-video model. enable_model_cpu_offload(): Enables offloading of the model processes to the CPU to reduce the GPU load.TextToVideoIFInterpPipeline.from_pretrained(): Sets up a pretrained model to interpolate between frames in a video based on the input text.IFSuperResolutionPipeline.from_pretrained(): Creates a pipeline to enhance the image resolution using a pretrained super-resolution model.TextToVideoIFSuperResolutionPipeline_Cond.from_pretrained(): Loads a pretrained model for super-resolution video frames based on the input text.VideoToVideoSDPipeline.from_pretrained(): Loads a pretrained model for converting video formats.enable_vae_slicing(): Enables Variational Autoencoder (VAE) model slicing to optimize performance.infer(): Generates a video from a text prompt by encoding the text and creating keyframes from the encoded data.export_to_video(): Exports the final video frames to a video file. The generated video is saved to /outputs within your project directory.Set Up Show-1 as a System Service
Follow the steps below to create a new Show-1 system service to securely run and control the application on your server.
Create a new system service file.
console
Add the following contents to the file. Replace
linuxuser with your actual user and /home/linuxuser/Show-1/ with your project directory. systemd
Save and close the file.
Reload the systemd daemon to synchronize the service changes.
console
Enable the Show-1 system service file.
console
Start the Show-1 system service.
console
View the Show-1 system service status and verify that it's running.
console
Output:
Wait at least
15 minutes before testing access to the Show-1 application to allow the model download process to complete.Set Up Nginx as a Reverse Proxy To Enable Access to the Show-1 Application
Install Nginx on your server.
console
Create a new Nginx virtual host configuration
show-1conf. console
Add the following configurations to the file. Replace
show-1.example.com with your actual domain. nginx
Save and close the file.
Activate the new virtual host configuration.
console
Test the Nginx configuration for errors.
console
Restart Nginx to apply the configuration changes.
console
Allow the HTTP port
80 through the default UFW firewall. console
Reload the UFW table to apply changes.
console
Access the Show-1 Interface and Generate Videos
The Show-1 model uses Gradio to generate a web interface to run the model and generate videos. Follow the steps below to run the model and generate a sample video from a text prompt.
Access your Show-1 domain using a web browser such as Firefox.
Enter your text prompt in the Prompt field. For example,
a rocket firing up to the sky.
Generate a Show-1 Video from Text Prompts
Click Submit to start the video generation process and monitor the model processing time.
Click the Download symbol in the top right corner of the generated video to preview and download the file on your computer.

Download a Generated Show-1 Video
Verify that the model saves your generated videos to the
/output directory in your project files.Conclusion
You have deployed the Show-1 Video Diffusion Model to generate realistic videos using text inputs on a Vultr Cloud GPU server. You can modify the application code to securely generate videos on your server using Show-1 with the super-resolution DeepFloyd model. For more information, visit the Show 1 project page.
Tags:
Related Content
Comments
No comments yet.
0 likes · 13 views · Jun 19, 2025
Deployed Show-1 model on Vultr Cloud GPU to generate videos from text prompts.