Gymmie - AI Form Scanner

Your AI gym coach that watches every rep, fixes bad form and prevents injury

Comment 5

🏋️ Gymmie — AI Form Scanner

Inspiration

I've been going to the gym consistently for 1.5 years. No missed workouts. Strict diet. Progress tracking.

Yet… the results just didn't match the effort.

After months of frustration, I finally discovered the real issue: my exercise form was wrong. Not "slightly imperfect" — fundamentally inefficient and sometimes unsafe. I was doing the work, but my body wasn't getting the stimulus it needed.

That realization became the spark for Gymmie.

Most gym-goers don't have access to a personal trainer. Mirrors lie. YouTube tutorials don't adapt. And bad form compounds silently over time. I wanted a way for anyone to instantly understand what they're doing wrong — using just their phone camera.

What Gymmie Does

Gymmie is an AI-powered gym form scanner.

Users record a short workout clip, and the app:

  • Analyzes exercise technique
  • Scores the form objectively
  • Highlights what's correct
  • Clearly explains what's wrong
  • Flags unsafe movement patterns
  • Provides concise coaching cues for correction

The focus is form quality, not motivation or workout plans. Good results start with good movement.

How I Built It

Tech Stack:

  • Frontend: Expo + React Native
  • AI: Gemini models for multimodal video understanding

Core Logic:

  • Short video input (≤ 30 seconds)
  • Strict, bias-free form evaluation
  • Biomechanics-focused feedback

Design Philosophy:

  • Minimal UI
  • No distractions
  • Clear coaching, not encouragement fluff

The app intentionally does one thing well: Tell you whether your form is helping or hurting your progress.

What I Learned

  • Consistency without correctness is wasted effort
  • AI can act as a scalable coach when used responsibly
  • Being strict in feedback produces better outcomes than being positive
  • Real problems come from lived experience — not ideas on paper

I also learned how challenging it is to:

  • Extract structured insights from raw video
  • Avoid false positives in movement analysis
  • Design feedback that is actionable, not generic

Challenges Faced

  • Ensuring AI feedback is critical but fair
  • Avoiding overly optimistic scoring
  • Handling unclear camera angles and lighting
  • Balancing real-time performance with mobile constraints
  • Designing trust: users must believe the feedback

Why This Matters

Bad form doesn't just slow progress — it leads to injuries, frustration, and people quitting altogether.

If Gymmie can help someone realize earlier what I realized much later, then it's already doing its job.

+ 17 more
Share this project:

Updates