Upload a photo or video of your swimming stroke and get instant AI coaching feedback. StrokeLab uses pose detection to analyze your body alignment, arm entry angle, head position, and more — then gives you a score breakdown with personalized tips to improve. Works from a single photo. No underwater cameras, no wearables. Free to try — 3 analyses per month, no signup required.
No reviews yetBe the first to leave a review for StrokeLab
Maker
📌
Hi everyone! 👋
I'm a swimmer and developer who got frustrated with not being able to see my own stroke form. After months of experimenting with pose detection and AI, StrokeLab was born.
The tech stack is pretty fun:
• MediaPipe for real-time body pose detection
• LLM-powered coaching that translates joint angles into actionable swim tips
• Next.js + Vercel for a smooth web experience
The biggest challenge was getting the AI to understand swimming biomechanics — not just "your elbow is at 45°" but "your catch is too shallow, try entering your hand further forward."
We're offering free access (3 analyses/month) to the PH community. Would love to hear:
• What features would you add?
• Any swim strokes you'd like us to support?
Thanks for checking it out! 🏊♂️
Report
How accurate is the pose detection from just a regular above-water photo, especially for the kick and hip position that you cant really see well?
Report
Maker
@elatanaslan Honest answer: above-water photos work well for upper body analysis (arm position, head alignment, body rotation), but underwater elements like kick technique and hip depth are harder to capture from surface-level footage.
For the most accurate full-body analysis, we recommend underwater or side-angle videos when possible. That said, our AI can still infer a lot from above-water footage using biomechanical models — it just won't be as precise for kick-specific metrics.
We're exploring multi-angle analysis to address this limitation. Thanks for asking — it's a common question!
Report
How well does the pose detection hold up with busy pool backgrounds or other swimmers in frame? Curious if that throws off the scoring.
Report
Maker
@dtanta24444 That's a really good point! Our pose detection model is trained to focus on the primary swimmer in frame. For best results, we recommend uploading videos where the swimmer is clearly visible and relatively isolated.
In busy pool environments, the model may occasionally pick up other swimmers, which can affect accuracy. We're working on improved person-tracking to handle this better. For now, cropping the video to focus on your swimmer gives the most reliable results.
Thanks for the feedback — this is exactly the kind of real-world scenario we're optimizing for!
Report
Does the AI feedback account for different swimming strokes (freestyle vs butterfly vs breaststroke) or does it apply the same scoring criteria across all of them?
Report
Maker
@doanyaarnl8fcx Great question! StrokeLab currently provides detailed scoring for all four major strokes — freestyle, breaststroke, butterfly, and backstroke. Each stroke has its own tailored scoring criteria. For example, freestyle analysis focuses on body rotation and catch angle, while breaststroke emphasizes pull symmetry and kick timing. The AI adapts its feedback based on the stroke type it detects, so you get sport-specific coaching tips rather than generic ones.
Report
Tried it with a photo of my freestyle stroke and was surprised how detailed the arm entry feedback was. Would be cool to see the score over time so I can track changes.
Report
Finally tried this on a terrible selfie from last week's open water swim and it nailed my head position being too high. Wild that it picks up that much from one phone photo, the tips were actually useful too.
How accurate is the pose detection from just a regular above-water photo, especially for the kick and hip position that you cant really see well?
@elatanaslan Honest answer: above-water photos work well for upper body analysis (arm position, head alignment, body rotation), but underwater elements like kick technique and hip depth are harder to capture from surface-level footage.
For the most accurate full-body analysis, we recommend underwater or side-angle videos when possible. That said, our AI can still infer a lot from above-water footage using biomechanical models — it just won't be as precise for kick-specific metrics.
We're exploring multi-angle analysis to address this limitation. Thanks for asking — it's a common question!
How well does the pose detection hold up with busy pool backgrounds or other swimmers in frame? Curious if that throws off the scoring.
@dtanta24444 That's a really good point! Our pose detection model is trained to focus on the primary swimmer in frame. For best results, we recommend uploading videos where the swimmer is clearly visible and relatively isolated.
In busy pool environments, the model may occasionally pick up other swimmers, which can affect accuracy. We're working on improved person-tracking to handle this better. For now, cropping the video to focus on your swimmer gives the most reliable results.
Thanks for the feedback — this is exactly the kind of real-world scenario we're optimizing for!
Does the AI feedback account for different swimming strokes (freestyle vs butterfly vs breaststroke) or does it apply the same scoring criteria across all of them?
@doanyaarnl8fcx Great question! StrokeLab currently provides detailed scoring for all four major strokes — freestyle, breaststroke, butterfly, and backstroke. Each stroke has its own tailored scoring criteria. For example, freestyle analysis focuses on body rotation and catch angle, while breaststroke emphasizes pull symmetry and kick timing. The AI adapts its feedback based on the stroke type it detects, so you get sport-specific coaching tips rather than generic ones.
Tried it with a photo of my freestyle stroke and was surprised how detailed the arm entry feedback was. Would be cool to see the score over time so I can track changes.
Finally tried this on a terrible selfie from last week's open water swim and it nailed my head position being too high. Wild that it picks up that much from one phone photo, the tips were actually useful too.