DIY-ifiy

If you are the type of person is a maker, crafter, designer with a large library of assets inspiration we can help you turn your project into a real 3d printed items.

Comment

Inspiration

AI fashion tools often stop at producing attractive images. We wanted to go one step further and explore what happens after a designer finds a visual direction they love.

DIY-ifiy was inspired by a simple question: Can we turn fashion inspiration into something a person can actually build and wear?

Our goal was to create a bridge between visual ideation and physical fabrication. A designer should be able to upload an image or capture a look through video, explain what they want it to become, and receive structured wearable components that can continue into a real 3D design workflow.

What it does

DIY-ifiy transforms images, uploaded videos, and live camera frames into buildable fashion concepts.

A user provides reference media and describes their creative intent—for example, “turn this shape into a lightweight protective collar.” Gemini analyzes the reference and generates wearable primitives such as cuffs, collars, shells, panels, and accessories.

The user can review these primitives, select the most promising result, and send it to open source Open-Cascade-SAS/OCCT CAD development library for transforming the concept into a physical-design and then return a downloadable STEP file for further CAD editing, prototyping, or fabrication.

Instead of treating an AI-generated image as the final product, DIY-ifiy treats it as the beginning of the manufacturing process.

How we built it

We built the interface using Next.js, React, and TypeScript.

The pipeline consists of five stages:

  1. Input: Upload a PNG, JPEG, WebP, or MP4, choose a sample design, or capture a frame from a live camera.
  2. Signal: Use video as a visual input source for capturing garments, movement, silhouettes, and inspiration.
  3. Direction: Add a natural-language description of what the design should become.
  4. Translation: Use Google Gemini for multimodal understanding and generation of wearable primitives.
  5. Materialization: Send the selected primitive to the project code connected through its local MCP interface and expose the resulting STEP artifact.

We structured the Gemini output into primitives containing information such as the component’s name, target body region, visual representation, and confidence score. This gives the user an intermediate design they can inspect before starting the more expensive 3D-generation process.

We also implemented browser-based camera and microphone controls, live frame capture, media uploads, sample fixtures, pipeline status indicators, and STEP-file downloads. The video layer establishes the interface for integrating Vonage Video API sessions and collaborative live design experiences.

Challenges we ran into

The largest challenge was translating between several very different representations:

visual reference → creative intent → wearable primitive → hardware-generation request → 3D artifact

Generated images can look convincing while still being physically ambiguous. We needed to extract smaller, named components that could be reasoned about individually instead of asking the system to generate an entire complex outfit in one step.

Connecting a responsive web application to the project code was also challenging because physical-design generation can take much longer than a typical AI request. We had to make the pipeline’s status visible, handle failures, and return generated artifacts without making the experience feel like a black box.

Video created another challenge. Uploaded videos, live

Built With

Share this project:

Updates