Stable Diffusion User Guide

0. Prerequisites

• What is Stable Diffusion?

  • An open-source generative AI model for Text-to-Image and Image-to-Image tasks
  • Designed to run on personal computers equipped with GPUs by significantly reducing computing resource requirements
  • Can be installed and executed in a 'local environment' on a private PC, independent of an online connection
    
    

• Gcube Sign-up and Point Purchase

  • Sign up on the Gcube official website(http://gcube.ai)

  

  

  • Log in and access User Mode

  

  • Click the “Point Charge" menu on the left sidebar

  

  • Select the desired amount of points to refill, and Choose a payment method and click the “Payment” button to complete the transaction.

  

1. gcube Platform Stable Diffusion Workload Preparation

• Register a New Workload

  • After logging in, click the “Create New Workload” menu on the left sidebar

  

• • Workload Information: Stable Diffusion Setup
  • Description: Enter [ Stable Diffusion ] in the workload description field

  

  • Container: Configure the Container Image for Stable Diffusion
    • Storage Type: Docker Hub
    • Container Image : [ universonic/stable-diffusion-webui ]
      • Click the "Verify Image" button next to the input field. Most of the time, the container port will be automatically populated
    • Container Port: 8080 (Automatically configured)

  

  • GPU Selection : Configure the Execution Environment
    • Target Tier: Tier 3 (Tier for individual users)
    • GPU Memory: 8GB
    • GPU : RTX 3070 X 1 10GB

  

  • Options (Optional): Detailed Container Settings
    • Replicas: Set the number of identical container instances to run in parallel
    • Minimum CUDA Version: Specify the minimum CUDA version required by the container
    • Shared Memory: Specify the size of the shared memory available to the container
    • ※ Note : For this guide, we will proceed without configuring any optional settings

  

  • Estimated Cost: Calculate the Expected Usage Fee
    • Click the checkbox next to “Immediate Deploy” to deploy immediately upon registration
    • Click the checkbox next to “Manual Deploy” to deploy manually after registration
    • Click the “Register” button to complete the workload registration

  

2. Running the Stable Diffusion Workload on Gcube Platform

• Post-Registration Screen

• Verify the information of the created workload

  • Click on “Stable Diffusion” to view details

  

  • Scroll to the bottom of the Workload Details screen to check the “Deploy Status”

  

  • Ensure the Pod status in the Deployment Status section is “Running”
  • Click the “Container Log” button

  

  • Installation is complete when you see the message shown in the image within the logs
  • ※ Note: Installation time may vary depending on container settings and hardware specifications

  

  • Return to the Workload Details screen and click the “Service URL Link”
  • ※ Note: If you click the link before confirming the container logs, the interface may not display correctly

  

  • If the screen below appears, Stable Diffusion has been successfully launched

  

3. Stable Diffusion Execution Example on gcube Platform

• Stable Diffusion Execution Example

  • Enter the prompt "A cat in a Hat" in the txt2img input field and click the "Generate" button

  

  • Confirm that an image similar to the one below is generated
  • ※ Note: Generated images may vary