Generative Design & AI Visualization Portfolio

Page 1


PORTFOLIO WANG WENHAO

Selected work from 2024 -2025

MLA, National University of Singapore (NUS)

Grasshopper & AI: Hybrid Workßow for Design

Architecture

made by Stable Diffusion

These are conceptual spatial renderings generated using Stable Diffusion, exploring ßuid, organic architectural forms with dynamic light and shadow interplay.

Architecture

made by Stable Diffusion

Using ControlNet to refine and adjust the composition while maintaining focus on the central elements of the generated architectural space.

Illustration

XYZ Multiple Testing

Trained an illustration-style LoRA model and conducted multiple tests to explore its adaptability across different sampling methods and scene compositions.

Lora training 03

Comparison of different LoRA models and their performance at varying weight values, analyzing the impact on detail, lighting, and architectural representation.

Parametric Retaining Wall & Rainfall Simulation

Simulation Video: https://youtu.be/7HabKgx8jhQ

1. Populate points on the surface based on a circular shape.

4. Calculate the Z-values in surrounding directions from the highest point, move one unit toward the lowest point, determine proximity to the retaining wall.

5. Use Anemone to loop this process and visualize the water ßow.

2. Generate retaining walls based on height.

Succession Simulation 05

Simulation Video: https://youtu.be/3_lV0VGce-Q

1. By setting the original cell condition, the site's boundary is assigned 'alive' information as the starting point of primary succession.

2. Subsequently, the proximity of alive cells is evaluated—if a dead cell is surrounded by more than three, it will die. New areas will be activated if they have between 0 to 7 neighboring cells.

Cellular automata systems are discrete computational systems that rely on a series of rules deÞned by a neighbourhood of cells. This project transform this system into an Ecological Succession Simulation.

Text-to-3D Exploration with ComfyUI

1. Text: outdoor,plant,potted_plant,traditional_media,painting_\(medium \),tree,watercolor_\(medium\),colored_pencil_\(medium\),no_ humans,scenery,grass,palm_tree,day,graphite_\(medium\),house,le af,Negative prompt: (worst quality:2),(low quality:2),(normal quality:2) ,lowres,(monochrome),(grayscale),blurry,signature,drawing,sketch,te xt,word,logo,cropped,out of frame

(Text-to-3D Workßow) Generate an image from text, then create a depth map, convert it into a point cloud, and Þnally model it in Rhino.
3. Depth map extracted from the image
4. Using a GH script to reverse-calculate point cloud height from depth map RGB and generate a 3D mesh model
2. Image generated by text

Parametric&ComfyUITerracing Visuals

1. Ridge Line Control for Terrace Formation
Photo
2. ConÞrm the Form
3. Segment layer without trees4. Segment layer with trees5. Depth layer

Dogecoin Visualization & Quantitative Testing in GH

Simulation Video: https://youtu.be/z61Q4-ev0zc

The ideal investment return rate is 3325%. According to the dollar-cost averaging (DCA) calculation, when the daily decline exceeds 5%, an investment of $656 would yield a return of 242%.

1. Convert historical Dogecoin data into candlestick chart format (high, low, open, close)

2. Visualization of weekly candlestick charts

3. Set the ratio to determine the buy/sell amount based on price ßuctuations. Use Mass Addition to calculate the total expenditure. Compare it with the Þnal day's holdings data to obtain the total proÞt. Run Galapagos to Þnd the optimal investment strategy.

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.