Therapix AI

Inspiration

Six months ago, I experienced a minor shoulder subluxation. During recovery, I was prescribed physiotherapy exercises to do at home. But once I got back, I quickly realized how difficult it was to remember each step. The handout I received had vague instructions, and some exercises weren’t even listed.

When I spoke to physiotherapists, I learned this wasn’t just my issue — many didn’t have access to effective tools to design and distribute personalized exercise plans. Existing platforms were limited, with outdated libraries (~3,000 videos) that barely scratched the surface of what’s needed.

This experience revealed a clear gap: therapists need more powerful, flexible tools to design, adapt, and share personalized routines — tailored for different injuries, age groups, and ranges of motion.

What It Does

Therapix AI is an intelligent exercise design platform powered by generative AI. It enables physiotherapists to:

• 🎥 Upload a reference video and enter a description or prompt

• 🧠 Generate custom, multi-angle demonstration videos

• 🛠️ Customize by injury type, age group, difficulty, and target area

• 🔍 Review & refine the generated output

• 📤 Share structured exercise plans seamlessly with patients

Therapists can start from scratch or modify existing exercises — all within a guided workflow that supports creativity and clinical needs.

Challenges We Faced

• Designing an intuitive UI that balances structure and flexibility

• Ensuring clinical accuracy in generative outputs

• Adapting to diverse patient needs (e.g., post-op, geriatric)

• Managing video rendering speed while maintaining high quality

• Building trust in AI-generated content among healthcare professionals

• Sora model is still in preview thus we were not able to fine tune the model to our needs and had to use query optimization to get consistent videos

Accomplishments We're Proud Of

• ✅ A functional prototype that generates AI-powered videos from therapist-defined inputs

• 🧭 A streamlined end-to-end workflow: from video upload to patient-ready output

• 📱 A patient interface with clear visuals and instructions for at-home compliance

What We Learned

• Step functions to help orchestrate tasks for us in the background without human intervention

• How to work in parallel with design team to build a system that is functional and usable