DES 212 · Design Media II
Home › Tutorial 2.1: AI for Visualization

Tutorial 2.1: AI for Visualization

Goal: Learn to communicate with AI and guide it to complete tasks controllably.

Part 1: Text-to-Image Generation

Instructions:

  1. Follow the tutorial on the web interface linked below
  2. After setting up Gemini Nana Banana, navigate through all the examples first
  3. When you find examples that inspire you, copy and paste the image(s) and text prompt into your Nana Banana interface to experience

GenAI Visual Examples(https://ubcdesignmedia.notion.site/)

Tips:

  • Write clear prompts: Be specific and detailed. Use LLMs (Gemini, GPT, etc.) to help refine your prompts and add details. You can also upload reference images and ask LLMs to describe them.
  • Start strong: The first prompt is very important. Describe your goal in detail from the beginning. Consider using a chatbot to help construct your initial prompt.
  • Use references: Upload reference images and ask a chatbot to describe the style, geometries, lighting, and other visual elements to get inspiration for your prompts.
  • No complex structures needed: You don't need to follow specific prompt engineering structures. Current LLMs can understand natural language prompts well.
  • Stay current: AI technology changes rapidly. When searching for resources or tutorials online, use only materials created within the last 6 months. This tutorial was prepared for February 2026 and will be updated regularly.
  • Image Resolution: Images generated by Nano Banana have limited resolution, so you'll need to upscale them. This page has some websites for image upscaling. You don't need to upscale for this exercise.

Submission Requirements:

Generate your own images by preparing your own input images and text prompts. You must explore 5 different use cases.

Submit the following:

  • A .doc or .pdf file containing:
    • Input images
    • Text prompts used
    • Output images generated
    • Clear sequence showing the relationship between inputs and outputs (formatting doesn't need to be fancy, just legible)
  • All generated images as separate files

Part 2: Image-to-Mesh Generation

Instructions:

  1. Prepare your image: You may need to run the Photo to 3D Building Model process first to clean up your image
  2. Generate the mesh: Follow the Image-to-Mesh generation process detailed on this page.

Submission Requirements:

Submit one screenshot of your generated mesh model (you do not need to download the actual mesh file).

Part 3: Node-Based System (Optional Reference)

Note: This section is optional. No submission required.

What is ComfyUI?

ComfyUI is a professional, node-based interface for AI generation workflows. It offers full control over every step of the generation process through a visual node system. You can create, save, and share custom workflow templates that chain together multiple AI operations—including text-to-image, text-to-video, image-to-video, image-to-mesh, and more. This makes it ideal for users who want precise control and reproducibility in their AI generation pipelines.

Options for Using ComfyUI:

  • Local installation: ComfyUI on GitHub - Open source and free. Download and run on your machine for full control and no usage limits.
  • Cloud-based: ComfyUI Cloud - Requires a subscription but eliminates the need for local installation and hardware requirements.

Tutorial Resources:

 

Submit on: Exercise 2.1

before 11:59pm Feb 13