Filling Pieces

Footwear Design with AI
Introduction
This project explores the integration of generative artificial intelligence (AI) into footwear design, aiming to bridge the gap between traditional methods and innovative approaches. Conducted in collaboration with Filling Pieces, the project focuses on enhancing the efficiency, creativity, and overall quality of the footwear design process using AI tools. Additionally, I addressed the constand need for innovation and the long time-to-market of footwear, while designing for technical minimalism.
Design brief
The challenge was to create an AI-oriented footwear design framework for Filling Pieces, aiming to address key challenges in their design process, such as enhancing creativity and speeing up the ideation phase. However, generic AI models often fail to capture the specific brand identity of brands like Filling Pieces. To overcome this, the project focused on fine tuning AI models to align with the brand’s unique aesthetic, ensuring that the generated designs were innovative and reflected the brand identity.
In this project, I developed and refined an AI-driven design framework while also documenting the entire process in a detailed report, to conclude my Master's degree in Integrated Product Design and the Delft University of Technology. The full version of this report can be accessed by clicking the button below.

The project involved extensive research, iterative development, and the practical application of AI tools to demonstrate their effectiveness in real-world design scenarios. Iterative cycles of ideation, creation, and validation were conducted, using tools like Stable Diffusion and LoRA fine-tuning to generate and refine design concepts.
Read full report
Project process
Footwear Production Graduation Internship at Filling Pieces
Feb 2024 — August 2024

Solution: New Style - Fuse Tech

To generate designs for new styles, I developed a workflow using LoRA models, which I trained specifically to capture Filling Pieces' brand identity and the style of the current seasonal theme. By incorporating reference images, the AI-generated designs achieved greater consistency and aligned more closely with the desired aesthetic, producing outputs that were both creative and accurate. This AI-oriented approach significantly accelerated the design generation process.

LoRA Models to capture Modern Architecture

Next to the LoRA models used for caturing the brand identity, the seasonal theme 'Modern Architecture' also needed to be captured. The first moodboard showcases this in essence, featuring contemporary buildings with clean lines, minimalist forms, and innovative use of materials. While this LoRA model captured the style, only using this influence led to abstract and distorted outputs, that were not recognizable as shoes anymore. Therefor, the strength could not be too high and an additional LoRA model was necessary.

The second moodboard highlights footwear inspired by modern architecture, characterized by sleek, geometric shapes, bold accents, and minimalism. Using this LoRA model balanced architectural inspiration with the identity of footwear, maintaining recognizable shoe forms and blending both design languages cohesively. The influence of both LoRA models can be seen in the generated images, in both shapes and use of materials.

As the designer, the most important task was to intentionally curate the moodboards of the images that were going to be used as the training images, as these would essentially have the greatest visual impact on the outcomes. With LoRA models, you cannot use too many images, therefore each image should add something unique, and it's details will learned by the AI model and ultimately have impact on the output.

Solution: Updating Existing Style - Cruiser Lite

For updating existing styles, such as the Cruiser silhouette, I added a ControlNet model to maintain consistent linework. This made sure that the generated designs remained recognizable and true to the original silhouette. The strength of the ControlNet model was adjusted to balance innovation and creativity with recognizability, determining how much the final design deviated from the original silhouette.

Design Decisions - Technical Minimalism

Both designs feature a neutral cream base with pops of orange, highlighting technical aspects in the minimalist style. The cream palette ensures a clean, timeless look, and allows the technical details to stand out without distraction.

Functional Orange Accents
Cruiser Lite (left): The orange highlights the heel clip, that enhances stability and reduces lateral movement.
Fuse Tech (right): The shank is marked in oranges, that provides torsional resistance and arch support.

These designs strike a balance between form and function, where technical features are subtly integrated in the minimalistic design approach. The pops of orange not only enhance visual interest but also signal key performance elements, that shows how AI-driven design can merge innovation with style.

Conclusion

In addition to 2D design processes, I also explored further potential for innovation, by using 3D generative AI to create 3D models of the footwear concepts. The technology was however not advanced enough yet to be used in the development process, therefore the renders only served as a means to enhance visual communication.

This project demonstrated that integrating AI into the design process at Filling Pieces significantly enhanced creativity and efficiency, but also revealed limitations in clarity and production feasibility. To address these, AI-generated designs should be paired with detailed technical drawings for precision in manufacturing. This underscores the need for a complementary relationship between AI tools and human designers, leveraging both to achieve optimal results.

Next project  →