Freeflow

Inspiration

A recent gym injury made us rethink how hard it is to know if your form is actually right, so we built Freeflow to make pose practice safer, clearer, and more interactive. We wanted to make exercise practice feel more like a game than a checklist, so Freeflow turns pose repetition into something fast, visual, and satisfying.

What it does

Freeflow lets users pick a workout, set their height and pace, then use a camera to match saved pose references in real time.

How we built it

We built a Python backend with MediaPipe pose detection and pose normalization, plus a React/Vite frontend for exercise selection, setup, and workout flow.

Challenges we ran into

Getting pose matching to stay stable across different body sizes and camera angles was the hardest part, along with keeping the camera pipeline and UI feedback in sync.

Accomplishments that we're proud of

We turned raw pose tracking into a usable training loop with saved reference poses, height-aware matching, and a clean workout selection experience.

What we learned

We learned how much normalization matters in computer vision, and how small UI details make pose feedback feel understandable instead of noisy.

What's next for Freeflow

Next, we want to build a mobile-first interface and run our inference model locally. We also want to expand the exercise library, improve scoring and feedback, make camera streaming smoother, and add more polished workout modes.