Camera Math Solver

What it is

Point your camera at a printed math problem, and it gets read, solved, and shown right on screen -- with nothing sent over the internet. It handles everyday arithmetic (like 45 + 55) and simple algebra with one unknown (like 2x + 3 = 11), and it's built to honestly say "I can't solve this" instead of guessing when a problem goes beyond what it's designed for. A "Freeze frame" button takes a clearer, closer look when glare or a shaky hand is making the live view harder to read.

There's also an experimental mode for handwriting, which switches to a different tool built specifically for reading handwritten text -- since tools built for printed text generally don't handle handwriting well.

Why everything happens on your device

Every interactive project on this site works entirely in your browser -- Narrate and Air-gapped Transfer do too. A camera feed is about as personal as it gets, so nothing here was ever going to be sent to a server: the reading and the math both happen right on your own device, in the same tab that's using your camera. Nothing your camera sees is ever sent anywhere.

Solving the math reliably

Rather than having AI guess at an answer, the actual math is worked out with simple, dependable calculation -- the same kind of process a calculator uses, not a guess. That means the answer is never "probably right" -- it's either correct, or an honest "I can't solve this one." Nothing in between.

What reading real handwriting and print actually looks like

Testing this against real camera photos (not perfectly clean text) turned up some genuine, everyday quirks in how text-reading tools behave:

• The multiplication symbol (×) and the letter "x" can look nearly identical. So this tool always treats a lowercase "x" as the unknown in an equation, and the true multiplication symbol as multiplication -- rather than guessing which one was meant.
• Small marks, like an exponent symbol, can get missed entirely. When that happens, the tool honestly says it can't solve the problem instead of guessing what was probably meant.
• Some characters can blend together, like a division symbol accidentally merging into a nearby number. Reading text from a photo is never perfect, and this tool is upfront about that instead of pretending otherwise.

None of this is a flaw in the math itself -- the calculation is always accurate once the problem is read correctly. These are simply the real, observed quirks of reading text from a photograph, and being honest about them beats quietly guessing and sometimes getting it wrong.

Making sure a reading is actually right

A single glance from the camera isn't trusted right away. The tool has to "see" the same thing twice in a row before it treats it as a real answer -- the same idea used in Air-gapped Transfer to avoid trusting a single blurry camera frame.

Why handwriting needed a completely different approach

Tools that read printed text well generally don't work for handwriting -- they're simply built and trained for different jobs. Testing that firsthand, rather than assuming it would just work, confirmed it: no amount of adjusting made a printed-text tool handle handwriting well. So handwriting mode uses a separate, purpose-built tool instead, switchable with a simple toggle.

Honest limitations

• Handwriting mode is new and hasn't been fully tested in everyday, real-world conditions yet -- the pieces are real and working, but its accuracy on real handwriting still needs a final check.
• Handwriting mode only works on a single frozen photo at a time (not a live, continuous scan), and works best with one line of writing framed clearly.
• On purpose, this only handles everyday arithmetic and simple one-unknown algebra -- more advanced math is out of scope by design, not because it was overlooked.
• Needs decent lighting and a clear, steady shot -- a page held at a sharp angle or in dim light will be harder to read.
• The character mix-ups described above are real, observed behavior from testing against real images, not a hypothetical.