TailorAI

Inspiration

I was inspired by the gap between AI-generated fashion imagery and real-world clothing. Many tools can generate visually appealing outfits, but they stop at images and ignore structure, measurements, and how a garment might actually be constructed.

I wanted to explore a different direction: instead of asking AI to generate finished clothes, could it help reason about garment structure—even in a simplified, imperfect way—and move one step closer to something sewable?

Pet clothing, and hoodies specifically, offered a focused scope to explore this idea without overgeneralizing too early. The same structural approach is intended to extend to more garment types and eventually human apparel as the system matures.

---

What it does

TailorAI is a structured, AI-assisted garment editor that converts outfit ideas into rudimentary, measurement-aware garment representations.

The system guides users through a step-by-step workflow:

• selecting garment structure (currently hoodie-only)
• defining body measurements
• adjusting visual and material properties
• previewing the design in 2D and 3D
• exporting an exploratory, pattern-like layout

While the long-term goal is sewable, fabrication-aware outputs, the current prototype produces approximate, schema-driven layouts that are not yet validated for professional sewing or manufacturing. While the current prototype focuses on pet hoodies, the underlying structure is designed to support additional garment categories and human measurements in future iterations.

---

How we built it

The app is built as a multi-step, stateful editor rather than a single AI prompt.

Core garment concepts—such as structure, measurements, and visuals—are modeled explicitly in code using constrained schemas. Gemini is used to assist with reasoning and suggestion within those constraints, rather than generating free-form outputs.

The frontend is implemented in React with a clear workflow:

1. Input (text, images, recycled materials)
2. Structural garment definition
3. Measurement editing
4. Visual refinement and AI-assisted edits
5. Export of a pattern-style layout with optional seam allowance

This approach emphasizes structure and workflow as a foundation for future sewable designs.

---

Challenges we ran into

Clothing is a physical domain where precision matters, but AI models and hackathon-scale systems are not inherently aware of real-world tailoring rules.

One challenge was defining how much structure to include without overclaiming accuracy. Another was limiting scope responsibly—supporting only hoodies—so the system could remain coherent and understandable.

Balancing visual appeal with technical honesty was an ongoing design consideration.

---

Accomplishments that we're proud of

• Built a fully functional, multi-step garment editor rather than a static demo
• Modeled garment structure and measurements explicitly instead of relying on image generation alone
• Integrated AI as a constrained assistant rather than an unconstrained generator
• Defined a clear direction toward sewable outputs while remaining honest about current limitations

---

What we learned

We learned that applying AI to physical-world domains requires clear boundaries and careful framing.

Even a simplified, unverified structure can be valuable when it establishes a path toward real-world making.

Built With

• gemini-3-api-(vision-+-text)
• html5
• react
• svg
• tailwind-css
• typescript
• vite