LiftIQ

Turn your camera into a real-time AI personal trainer

Comment

Inspiration

Many people go to the gym or work out at home, but they are not sure if they are doing exercises correctly. Bad form can make workouts less effective and can even cause injuries. Personal trainers can help, but they are expensive and not always available. We wanted to build something simple that can help anyone improve their form using just a camera.

What it does

LiftIQ is an AI-powered workout coach that uses your camera to track your body in real time. It can recognize exercises like squats, push-ups, and lunges. It counts reps, gives a score out of 100, and shows what you are doing wrong. It also gives live feedback like “go lower” or “keep your back straight,” and highlights joints that need correction. After the workout, it shows a summary with your performance, mistakes, and progress.

How we built it

We built LiftIQ using modern web technologies. The frontend is built with Next.js, React, TypeScript, and Tailwind CSS, with Framer Motion for smooth animations and Radix UI for accessible components. For real-time body tracking, we use MediaPipe Pose Landmarker running directly in the browser. We created a custom angle-cycle state machine to accurately count reps and detect form issues. Voice coaching is powered by the browser's SpeechSynthesis API, and the Gemini API generates plain-English explanations of form mistakes. The backend uses Supabase for authentication, database, and cloud storage, with Zustand for client-side state management. Nutrition tracking integrates the USDA FoodData Central API, and workout recordings are stored locally in IndexedDB with automatic cloud sync. Recharts powers the analytics dashboard for visualizing progress over time.

Challenges we ran into

One of the biggest challenges was achieving high accuracy in rep counting and form detection. We wanted the system to feel reliable in real-time, but getting it close to 100% accuracy was difficult. Small changes in movement or camera angles could affect detection, so we had to carefully design our logic to handle different cases. We focused on making the system consistent and stable so it works well during real workouts.

Accomplishments that we're proud of

We are proud that we were able to build a fully working real-time AI workout coach within a short time. The system can track body movement, count reps, and give feedback while the user is exercising. We are especially proud of improving the accuracy of rep counting and making it stable enough to work during live demos. We also built a voice coaching system that gives helpful feedback without being annoying. Another thing we are proud of is the overall user experience. We designed the app to be simple, clean, and easy to understand, while still showing useful insights like scores and progress. Overall, we are proud that we turned a complex idea into a working product that feels like a real fitness tool.

What we learned

We learned that building accurate real-time AI systems is challenging. Even small errors can affect the user experience. We also learned that making the system stable and easy to use is more important than making it overly complex.

What's next for LiftIQ

In the future, we want to support more exercises as well as have more features such as detecting the amount of calories in a food item using just a picture. We also want to add personalized workout plans and smarter coaching features.

Built With

Share this project:

Updates