AI x Design

Empowering artists and hobbyists with powerful tools to realize their creative vision

AI design tools are an opportunity to democratize design and development, making the process more accessible and inclusive. Generative AI is an exciting new medium for creative expression, akin to any tool in an artist’s kit: a means for human expression driven by the creator's imagination.

In April 2023, Maison Meta launched the first ever AI Fashion Week, featuring AI-generated collections from global participants. An effort to support emerging designers working with the technology. Winners included José Sobral, Matilde Mariano, and Opé Majek - none of whom were trained designers. The winning collections were developed and sold by Revolve.

At present why is design not an inclusive endeavour :

1. Prototyping is costly which limits creative explorations for small businesses and individual artists. Simulating design, pattern, and fabric behaviour virtually can allow seeing the end product before prototyping. This saves time, reduces waste, and cuts costs. According to Rasberry AI, this can lead to “3-5 Hours saved on each design and up to 30% savings on sample cost”.

2. Even to use current design simulation softwares (which have their own limitations - see below) requires costly hardware and a steep learning curve. This makes onboarding a time and financial investment before have a clear value in sight when starting.

Recent developments in Gen AI may offer solutions that allow generating designs using detailed prompts on cloud and empower more people to seamlessly use tools without dedicated training to create and test designs virtually.

Disclaimer: Concerns with Gen AI tools in the creative process

Last year I attended a conversation in LA hosted by Will.I.Am, where musicians expressed concerns about AI, particularly the lack of clarity around companies using intellectual property in model training and its impact on royalties. As we navigate solutions, many independent bodies are promoting art and design using AI and even pushing them into retail.

In order to start the process of regulating the use of AI in the EU, across industries like Fashion, the European Parliament recently passed the AI Act. The Act categorizes AI systems by risk level, mandates transparency in AI training data, and ensures ethical practices, highlighting the need for similar regulations globally. These measures start the important conversations around protection of intellectual property and consent of use in training data.

Companies working on Gen AI art

Many companies are pushing the frontier on AI Art Generation. Some use cases include:

• Text to Image (DALL-E, Midjourney, Stable Diffusion)

• Text to Video (LUMA AI, Runway)

• Image to Video (LUMA AI)

Common use cases designers use tools for :

• Simulate design and fabric properties (Adobe Firefly, Metail)

• Create 3D garment from 2D (CLO)

These tools have several limitations which Generative AI can help bridge :

• Fit : Capturing the exact fit and drape of garments on different body types can be challenging. Simulations may not accurately represent how a garment will fit on diverse body shapes, leading to potential discrepancies between virtual and actual fits.

• Fabric simulation : Simulating how different fabrics move, stretch, and interact with the body is complex. Current tools struggle to accurately replicate the nuanced behaviors of various fabrics, such as the elasticity of knits or the drape of silk.

• Realism of Material Textures: Rendering textures and details like stitching or embellishments accurately. Some tools may not fully capture the tactile qualities of materials, impacting the realism of the simulation.

• Complexity for Beginners: These tools often require significant learning and practice to use effectively. Beginners may find the interfaces complex and the learning curve steep, which can be a barrier to entry.

• User Experience: Intuitive design and ease of use are crucial for efficient workflow. Some tools may have non-intuitive interfaces that slow down the design process.

• Hardware and Computational Requirements: 3D simulations require substantial computing power and memory. Users with less powerful hardware may experience slow performance or may be unable to run the software effectively. High-quality simulations can take considerable time to render. Long rendering times can delay the design process, especially for complex garments or detailed textures.

Advances in generative AI will make 3D content more realistic, accessible, and efficient, enhancing fabric simulation across diverse fabrics, body shapes and environmental factors.

TDLR:

DALL-E 2 (OpenAI)

• Text description to high-quality image

Midjourney

• Text description to high-quality image

• Users can iteratively refine generated images by tweaking prompts allowing for a highly customizable creative process

Runway

• Text description to video generation. New release: Gen-3 Alpha offers improved realism and extended length of clip form 2-3 to 10-11s

• Additional tools include : frame interpolation, upscale image

• Fashion designers use Runway to explore new concepts, generate patterns, and visualize garments in different styles and materials, with AI offering real-time feedback for efficient concept iteration.

Stable Diffusion (Stability AI)

• Text description to high-quality image

Adobe Firefly

• AI-driven tools for creating custom patterns, generating design variations, and automating tasks like color matching and fabric simulation, widely used in fashion design.

CLO Virtual Fashion

• Creates realistic 3D garments from 2D patterns, allowing designers to visualize, modify, and simulate their designs on virtual models, significantly speeding up the design process.

VStitcher (Browzwear)

• Employs AI to enable virtual garment creation and simulation, which helps fashion designers to quickly iterate on designs and assess the fit and drape of garments digitally.

Metail

• Uses AI to create customizable digital garments, enabling fashion designers to visualize how clothes will fit and look on different body types, thus improving the design and fitting process.

Luma AI

• Specializes creating realistic 3D models from photos, videos