AI Speed Cube Trainer
A free speedcubing recognition tool and trainer.
Tap the pattern you see on your physical cube and instantly identify the case, get the optimal algorithm for your holding angle, and track your progress over time. No account required, just open and train!
The recognition bottleneck
Two cubers. Same algorithms.
Same turning speed.
Seconds apart.
The faster cuber isn't moving their fingers faster. They're seeing the case faster. The pause between finishing F2L and starting OLL is where time disappears: not in the execution, but in the recognition gap.
Where time actually dies
The algorithm takes 1.5 seconds. The hesitation takes 3.
Most intermediate cubers plateau around 20 seconds because they've memorized the algorithms but not the patterns. The seconds you get back come from cutting recognition time, not execution time.
What we built
A trainer that drills the reflex, not the recipe.
Instead of flipping through algorithm sheets, you tap what you see and the trainer identifies the exact case, checking all four holding angles automatically.
How it works
Pattern detection across all 4 AUF
rotations, instantly.
For OLL, you mirror the sticker pattern from your physical cube onto the tappable grid. The engine matches it against all 57 OLL cases across every holding angle and tells you the angle to start from. For PLL, you draw arrows showing which pieces need to swap. No rotating the cube to match a diagram.
Tap what you see
Hold your cube. Tap the stickers you actually see.
Open the OLL detector. The tappable grid shows the top face and the four side strips. Tap each colored square to match your physical cube right now, or load the sample Sune pattern to see a match immediately.
The trainer identifies
Press DETECT MATCH. The case appears. The angle is shown.
The trainer identifies your exact OLL or PLL case, gives you the optimal algorithm, and shows which direction to hold your cube before you start. No squinting at tiny reference images.
Drill and track
Rate yourself. Hard cases come back. Streaks and mastery tracked for you.
The F2L flashcard trainer adapts to you. Cases you miss come back more often until they're mastered. Your streak and mastered count update in real time, saved locally in your browser.
What you get
Faster recognition. Lower times.
Every single solve.
Fifteen focused minutes a day of deliberate recognition drilling builds the reflex that separates intermediate from advanced cubers, across the complete OLL and PLL case sets, plus the core F2L patterns.
No catch
Free forever. No account.
No email.
Privacy-first.
Your practice progress lives in your browser: score, display name, favorites, and F2L mastery. No account, no email, nothing uploaded. Works on any phone, tablet, or desktop. Built by a speedcuber who got tired of flipping through printed algorithm sheets.
Try it now: OLL recognition
Open your cube. Finish F2L. Tap what you see.
The OLL detector is the fastest place to start. Tap the pattern on the top face and side strips, press DETECT MATCH, and the trainer identifies which of the 57 OLL cases you're looking at, and exactly which angle to hold. No login. No setup.
Start here
What recognition training actually is
Most intermediate speedcubers can already turn fast enough. The pauses that keep averages stuck in the mid-20s are almost always recognition: finishing F2L, staring at the last layer, and matching a pattern to an algorithm from memory. Recognition training is deliberate practice for that pause. You look at a case (on a physical cube or a diagram), name it, then check yourself.
AI Speed Cube Trainer is a free, browser-based tool for that loop. For OLL and PLL, you tap the sticker pattern you see on your physical cube. The site matches it against the full case set (57 OLL, 21 PLL), including how far the cube is rotated, and shows algorithms for that angle. For F2L, a spaced-repetition flashcard deck of 20 core pair cases saves mastery progress in this browser. There is no account and no upload of your practice data.
The site is built and maintained by Trevor, a hobbyist cuber (cubing since 2016, personal best 3x3 single 18.2 seconds, not a WCA competitor). The guides are written from the same plateaus most intermediate cubers hit. Start with the trainer if you already know CFOP, or with a guide if you are still building foundations.
Unlike a static algorithm wiki, the trainer forces active recall: you map a real pattern, check the answer, and repeat. That is the mechanism that shortens last-layer pauses. Favorites and a local practice score help you track volume without creating an account. Clearing site data in your browser resets progress, which is intentional for privacy.
Open the trainer
- OLL pattern detector — tap the top face and side strips, then DETECT MATCH
- PLL arrow detector — draw swaps and cycles on the top face
- F2L flashcard trainer — rate Got It / Hard / Missed; progress is saved locally
- How to use the trainer — step-by-step walkthrough of each mode
Guides and reference
Who this helps
If you average roughly 20–30 seconds and your last layer still freezes, you are the primary audience. Beginners who just finished the beginner method can use the CFOP guide and F2L flashcards before full OLL. Advanced hobbyists can use the detectors as a short warm-up before timed home sessions. Coaches and clubs can hand students a link with no install and no account.
For a first-person account of the intermediate plateau (six months stuck near 25 seconds until recognition practice changed things), read Why I Was Stuck at 25 Seconds. For hardware and setup, start with Choosing Your First Speedcube and the Cube Setup Guide.
About this project
This is an independent educational site, not affiliated with the World Cube Association or Rubik's Brand Ltd. Algorithms referenced here are standard community knowledge; the training interface, writing, and diagrams are original to this project. The product does what it claims: pattern matching for OLL and PLL, spaced-repetition F2L flashcards, and local-only practice progress. Read the about page for the full origin story, or contact Trevor with algorithm corrections or product feedback.
Privacy and terms: Privacy Policy · Terms of Use