FitMotion AI Review
A Python + Streamlit demo for exercise recognition and rep counting from video or webcam, with an optional OpenAI chatbot.
Each review covers deployability, value versus commercial alternatives, and privacy model. Tools that can run locally were started and exercised; mobile or backend-dependent tools were assessed from published builds, source code, and deploy guides. Ratings reflect what we were able to verify.
FitMotion AI — Open Source Exercise Recognition Tool
A Python + Streamlit demo for exercise recognition and rep counting from video or webcam, with an optional OpenAI chatbot. Last updated: 2026-06-21.
One-sentence verdict: A functional computer-vision prototype for a handful of exercises, useful for developers exploring local pose estimation but not a polished daily trainer.
What the System Is
FitMotion AI is a Streamlit app built around Python 3.10, TensorFlow, Keras, MediaPipe, and OpenCV. The documented capabilities include:
- Exercise recognition for Bicep Curl, Push Up, Squat, and Shoulder Press.
- Rep counting from an uploaded video or live webcam feed.
- Pose analysis using MediaPipe landmarks.
- Optional AI chatbot powered by LangChain + OpenAI API.
The chatbot requires a separate OpenAI API key; the rest of the app runs locally. The deploy guide notes that requirements.txt is incomplete and one deprecated Streamlit option must be removed to run.
| Key data | |
|---|---|
| Category | Fitness & Movement |
| Language | Python / Streamlit |
| License | Unknown |
| Self-hosted | Yes |
| AI provider | Local MediaPipe + optional OpenAI GPT |
| Dependencies | TensorFlow, Keras, MediaPipe, OpenCV, scikit-learn |
How to Install and Deploy
The deploy guide recommends a Python 3.10 environment and notes that extra packages are required beyond what requirements.txt lists.
cd /data2/docker/going-global/repos/Fitness-AI-Trainer-With-Automatic-Exercise-Recognition-and-Counting
/root/miniconda3/envs/metagpt_env/bin/python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pip install streamlit langchain langchain-community python-dotenv
# Remove the deprecated st.set_option line from main.py if startup fails
export OPENAI_API_KEY=sk-... # only needed for the chatbot
streamlit run main.py --server.port 5016 --server.headless true
Then open http://localhost:5016.
How to Test
The documented test flow is:
- Open
http://localhost:5016. - Select an exercise from the dropdown (Bicep Curl, Push Up, Squat, or Shoulder Press).
- Upload a short video or enable the webcam.
- Start the analysis and verify that MediaPipe landmarks appear and the rep counter updates.
- Optionally, with an OpenAI API key exported, test the chatbot for basic fitness questions.
The app is resource-intensive and supports only four exercises. The UI is functional but not polished.
Privacy & Compliance
FitMotion AI is not HIPAA compliant. Video processing happens locally through MediaPipe and TensorFlow. The optional chatbot sends text to OpenAI. Do not upload video containing PHI, and do not use the app for clinical movement analysis.
FitMotion AI vs Commercial Fitness Apps
| Dimension | FitMotion AI | Commercial Trainer (e.g., Freeletics, Future) |
|---|---|---|
| Cost | Free / self-hosted | Subscription |
| Data model | Video processed locally | Cloud processed |
| Exercise library | 4 exercises | Hundreds |
| Real-time feedback | Webcam + rep counter | Often richer form coaching |
| Mobile apps | Browser only | Native iOS / Android |
| Setup effort | High: conda env, missing deps, deprecated option | Low: app store install |
| AI chat | Optional OpenAI | Built-in coaching |
Who Should Use It
- Developers exploring MediaPipe pose estimation and exercise counting.
- Researchers building custom fitness prototypes.
- Users comfortable with Python environments and dependency troubleshooting.
Who Shouldn't Use It
- People looking for a polished daily workout app.
- Users who want guided programs, social features, or wearable sync.
- Anyone needing clinical-grade movement analysis.
FAQ
Which exercises does FitMotion AI recognize?
Bicep Curl, Push Up, Squat, and Shoulder Press.
Does the video leave my machine?
No. Video processing and pose estimation run locally with MediaPipe and TensorFlow. Only the optional chatbot sends text to OpenAI.
Why do I need to install extra packages?
The checked-in requirements.txt does not list streamlit, langchain, or langchain-community. They must be installed manually to run the app.
Verdict
FitMotion AI is a functional computer-vision prototype for a handful of exercises. It proves the core pose-estimation and rep-counting pipeline. It is not a polished daily trainer, but it is a credible starting point for developers who want to build on top of MediaPipe.
Ratings: Deployability 2/5 · Value vs Commercial 3/5 · Privacy Compliance 3/5