Sketch2Circuit

Sketch2Circuit V2 is a browser-based engineering studio that turns Arduino/C++ code into a verified hardware design: a circuit schematic, an auto-routed PCB, and an interactive 3D digital twin. It then checks the real breadboard on your desk against that design. THE PROBLEM Students and makers write code in one place and wire hardware in another. Nothing connects the two, so a wrong pin or a missing resistor shows up as a dead LED or a burnt part, and beginners can't tell whether the bug is in the code or in the wires. WHAT IT DOES • AI Studio: paste any Arduino/C++ code. Our AST engine extracts pins and connections in real time and builds the circuit live, with no predefined templates. • 3D Digital Twin: interactive Arduino, breadboard and jumper wires (Three.js) with info callouts and Obsidian/Blueprint themes. • 2D Schematic & PCB: schematic view plus an auto-router that places parts on an Arduino Uno R3 shield outline and routes traces at 0.5 mm with live DRC. • Verify & ERC: 18 electrical rules and one-click Auto-Verify & Repair for code issues, pin conflicts and pull-up setup. • Scan Breadboard: upload a photo of your real circuit. A vision model reads it into structured JSON (components and connections), then our code compares it with the digital twin and reports specific, actionable mistakes. • Code from scan: Arduino code is generated from what was actually detected, with safety warnings (LED without resistor, shorted rails, D0/D1 used with Serial), and can be sent to the editor to update the 3D twin. • Simulation (virtual oscilloscope), BOM, component catalog, calculators and exports. DESIGN PRINCIPLE: AI PERCEIVES, CODE DECIDES The vision model only converts a messy photo into a fixed schema. Pass/fail, error messages and code generation are deterministic, so results are explainable and repeatable. Users can review and remove false detections before anything is generated. STACK React 19, TypeScript, Three.js, a custom SVG schematic renderer, and the Fireworks AI inference API with Meta's Muse Spark Open as the primary vision model and automatic fallback to other models. TRACTION The first demo passed 42,000 views and thousands of engineers asked for access. V2 is the full studio, built for the Meta AI Hackathon.