Modular GenAI Design Systems

Modular GenAI Design Systems

Modular GenAI Design Systems

The goal of each system is to give designers better starting points, faster.

[

[

SCROLL TO VIEW MORE

SCROLL TO VIEW MORE

SCROLL TO VIEW MORE

SCROLL TO VIEW MORE

SCROLL TO VIEW MORE

SCROLL TO VIEW MORE

]

]

[Breakdown of five AI systems that compress the distance between idea → artifact → athlete feedback]

[5 AI systems speed up workflow between idea → artifact → athlete feedback]

[Breakdown of five AI systems that compress the distance between idea → artifact → athlete feedback]

1. THE ATHLETE MOMENT system

1. THE ATHLETE system

1. THE ATHLETE MOMENT system

This framework says: “AI doesn’t invent shoes. Athletes do.”

This framework says: “AI doesn’t invent shoes. Athletes do.”

[Core Nike DNA]

[Core Nike DNA]

Concept

Concept

An AI system that generates footwear concepts starting from athlete context, not aesthetics.

An AI system that builds concepts from athlete context.

Inputs

Output

Prompt

Style Direction

Inputs

Output

Prompt

Style Direction

Inputs

Output

Prompt

Style Direction

step 1: generate athletic movement


  • Start with the athlete

  • Respect Nike’s visual heritage

  • Gather information on body mechanics, muscle engagement, and shoe usage

step 1: generate athletic movement

  • start with the athlete

  • respect Nike’s visual heritage

  • gather info on body, muscles, usage of shoe etc.

step 1: generate athletic movement


  • Start with the athlete

  • Respect Nike’s visual heritage

  • Gather information on body mechanics, muscle engagement, and shoe usage

Step 2: ATHLETE FORCE ANALYSIS (human lead not AI)

Step 2: ATHLETE FORCE ANALYSIS (human lead not AI)

Notes based on the AI imagery created above.
This part of the system is intentionally not AI. Assessment and analysis are performed by designers.

Notes based on the AI imagery created above.
This part of the system is intentionally not AI. Assessment and analysis are performed by designers.

[Design intent and logic]

Athlete Moment
Force
Body Area
Block exit
Forward propulsion
Forefoot
Toe-off
Max load
Metatarsals
Acceleration
Stability
Heel + midfoot

Step 3: FOOTWEAR CONCEPT 

Step 3: FOOTWEAR CONCEPT 

Generate the shoe as a response to the force logic. Then Human Critiques

Prompt:
Performance sprint footwear concept designed for explosive block starts. Aggressive forefoot stiffness informed by horizontal force vectors, sculptural carbon plate implied through form language, elevated toe spring, narrow locked midfoot, minimal stabilized heel. No decorative elements, no branding, no lifestyle cues. Form follows performance.

Generate the shoe as a response to the force logic. Then Human Critiques

Prompt:
Performance sprint footwear concept designed for explosive block starts. Aggressive forefoot stiffness informed by horizontal force vectors, sculptural carbon plate implied through form language, elevated toe spring, narrow locked midfoot, minimal stabilized heel. No decorative elements, no branding, no lifestyle cues. Form follows performance.

Step 4: Designer-Controlled Variations

Step 4: Designer-Controlled Variations

highlighting tool behavior. These workflows are built in WeavyAI or ComfyUI for precise control over structure, constraints, and iteration. Custom LoRA models encode footwear-specific form language, including sole geometry, toe spring behavior, and performance proportions. This keeps outputs grounded in functional design while allowing quick variation that maintains brand consistency.

tweak ONE variable each time:

  • Forefoot stiffness +/-

  • Heel width +/-

  • Toe spring angle +/-

  • Midsole thickness +/-

    Note: LoRA fine-tuning ensures variations explore structure, not style

highlighting tool behavior. These workflows are built in WeavyAI or ComfyUI for precise control over structure, constraints, and iteration. Custom LoRA models encode footwear-specific form language, including sole geometry, toe spring behavior, and performance proportions. This keeps outputs grounded in functional design while allowing quick variation that maintains brand consistency.

tweak ONE variable each time:

  • Forefoot stiffness +/-

  • Heel width +/-

  • Toe spring angle +/-

  • Midsole thickness +/-

    Note: LoRA fine-tuning ensures variations explore structure, not style

VARIABLE INPUT #1

RESULT

VARIABLE INPUT #2

RESULT #2

Step 5: Designer-Approved final Output

Step 5: Designer-Approved final Output

AFTER MULTIPLE VARIATIONS WE GET A FINAL SHoE DESIGN

AFTER MULTIPLE VARIATIONS WE GET A FINAL SHoE DESIGN

Step 6: PAIR shoe with athlete

Step 6: PAIR shoe with athlete

Athlete Image → Shoe Image


Footwear geometry driven by peak horizontal force during block exit.

Athlete Image → Shoe Image


Footwear geometry driven by peak horizontal force during block exit.

other generated sketches from workflow - plausible+implausible

Athlete Image → Shoe Image


Footwear geometry driven by peak horizontal force during block exit.

Athlete Image → Shoe Image


Footwear geometry driven by peak horizontal force during block exit.

[more to build]

Additional Systems in Development

Additional Systems in Development

Additional Systems in Development

02

THE DESIGNER CO-PILOT (seen above)

THE DESIGNER CO-PILOT (seen above)

Setup:

A ComfyUI / Weavy.AI workflow that responds to designer intent, not replaces it.

Visual Rule:

  • AI is never “finished” until a designer touches it.

Workflow:

  • Sketch → AI interpretation → designer correction → AI adapts (similar to above)

  • Visible feedback loops


Sequence:

  1. Human sketch (messy, imperfect)

  2. AI interpretation

  3. Designer override (circles, strike-throughs)

  4. Refined AI output that respects corrections


Why:
AI should never erases authorship.

Setup:

A ComfyUI / Weavy.AI workflow that responds to designer intent, not replaces it.

Visual Rule:

  • AI is never “finished” until a designer touches it.

Workflow:

  • Sketch → AI interpretation → designer correction → AI adapts (similar to above)

  • Visible feedback loops


Sequence:

  1. Human sketch (messy, imperfect)

  2. AI interpretation

  3. Designer override (circles, strike-throughs)

  4. Refined AI output that respects corrections


Why:
AI should never erases authorship.

03

FOOTWEAR DNA MODEL

FOOTWEAR DNA MODEL

Setup:

A fine-tuned model trained on:

  • Nike outsole logic

  • Midsole proportions

  • Toe spring rules

  • Heel geometry language


Visual Rule:

  • Focus on structure not on logos.

Workflow:

  • Generate based on logic input → output multiple brand dna models → Nike model clearly “feels” right without a Swoosh


why
This shows:

“Brand is geometry, not graphics.”

Setup:

A fine-tuned model trained on:

  • Nike outsole logic

  • Midsole proportions

  • Toe spring rules

  • Heel geometry language


Visual Rule:

  • Focus on structure not on logos.

Workflow:

  • Generate based on logic input → output multiple brand dna models → Nike model clearly “feels” right without a Swoosh


why
This shows:

“Brand is geometry, not graphics.”

04

PIXEL TO PROTOTYPE 

PIXEL TO PROTOTYPE 

Workflow:

  • AI → CAD → 3D print → physical object

  • This is an iterative heavy process


Sequence:

  1. AI concept render

  2. Raw 3D print (stringing, layer lines visible)

  3. Refined prototype


output:
Triptych layout showing failure, iterations and prototypes

Why:
speeds up the building process by 30%-45%


Workflow:

  • AI → CAD → 3D print → physical object

  • This is an iterative heavy process


Sequence:

  1. AI concept render

  2. Raw 3D print (stringing, layer lines visible)

  3. Refined prototype


output:
Triptych layout showing failure, iterations and prototypes

Why:
speeds up the building process by 30%-45%


05

FUTURE FOOTWEAR ATLAS

FUTURE FOOTWEAR ATLAS

Setup:

a Weavy.Ai + n8n + LLM system that stays up to date on how sport, materials, and culture are changing, and uses that information to continuously feed a generative design workflow so teams can explore ideas quickly and stay prepared.

A "living" AI model real time mapping:

  • Sports evolution

  • Material innovation

  • Cultural shifts

Output (visual language):

  • Archive meets future

  • Text overlays like Nike internal decks

  • Minimal color


Why:

prepreration beats preditcion.

Setup:

a Weavy.Ai + n8n + LLM system that stays up to date on how sport, materials, and culture are changing, and uses that information to continuously feed a generative design workflow so teams can explore ideas quickly and stay prepared.

A "living" AI model real time mapping:

  • Sports evolution

  • Material innovation

  • Cultural shifts

Output (visual language):

  • Archive meets future

  • Text overlays like Nike internal decks

  • Minimal color


Why:

prepreration beats preditcion.