Kinetiq

Inspiration

As college students finding our footing in the fitness world, we’ve all felt that hesitation at the gym, especially as women, when trying a new exercise and not being 100% sure if our form is correct. That uncertainty can be intimidating, and bad form not only reduces results but can also increase risk of injury. We also know many girls prefer working out at home and don’t have access to a coach for feedback. We wanted to build something that acts like a real-time personal trainer - accessible, private, and encouraging.

What it does

Kinetiq is a computer vision-powered form coach that uses your webcam to analyze your movement in real time. It tracks joint positions, calculates biomechanical angles, counts reps accurately using a movement state machine, and gives occasional voice corrections when your form breaks down. The goal is to provide actionable, real-time feedback, not just rep and set counting, so users can train safely and confidently.

How we built it

We used MediaPipe Pose to extract 33 body landmarks from live webcam footage in the browser. From those landmarks, we calculated joint angles (like shoulder-elbow-wrist) using vector math and trigonometry to determine movement phases (bottom vs. lockout).

We implemented a state-based rep counting system to avoid double-counting caused by angle jitter. For feedback, we added threshold-based posture checks (like shoulder symmetry) and triggered voice coaching using the browser’s SpeechSynthesis API with cooldown logic to prevent spam.

Everything runs client-side in JavaScript, which keeps it lightweight and accessible without needing a backend server.

Challenges we ran into

One major challenge was angle instability and rep miscounting from MediaPipe. Small tracking fluctuations would cause false transitions between movement states. We solved this by adding stricter angle thresholds and a controlled “down → up” state machine.

Another challenge was voice feedback - initially it overlapped or triggered too often. We implemented timing buffers and cooldown logic so corrections feel natural and supportive rather than overwhelming.

Accomplishments that we're proud of

We’re proud that we built a functional real-time form correction system in a weekend. It doesn’t just count reps - it evaluates posture and gives intelligent feedback.

We’re especially proud that the project solves a real problem we personally experience. It feels meaningful to create something that could genuinely help women feel more confident training, whether at the gym or at home.

What we learned

We learned how powerful browser-based computer vision has become - being able to run pose estimation in real time without a backend is incredible. We also learned that small implementation detail can make a huge difference in user experience.

What's next for Kinetiq

Next, we want to: -Expand to multiple exercises (squats, deadlifts, lunges, push-ups) -Add bilateral tracking for better symmetry detection -Incorporate tempo tracking and range-of-motion scoring -Build a progress dashboard to track improvement over time -Potentially integrate a lightweight AI model for personalized feedback patterns

In the long term, we envision Kinetiq becoming a smart, accessible coaching tool designed especially for beginners and women entering strength training.

Built With

• css
• html
• javascript
• mediapipe-pose
• speech-synthesis