From photo to solution —
here’s exactly what happens.
A complete walkthrough of every stage: how the photo is read, how the problem is interpreted, and how the step-by-step solution is built. No black box.
You upload a photo or type your problem
The starting point is whatever you have in front of you. A photo taken directly from your phone camera, a screenshot of a digital problem, or a typed question — all three work. The photo route is the most common because it’s fastest and handles things that text can’t, like diagrams and handwritten notation.
For photo uploads, the image is sent securely to the recognition layer. It isn’t stored on our servers after processing — once the problem is extracted from the image, the image itself is discarded. Nothing is retained or linked to your session.
- Frame just the problem — not the entire page
- Good lighting matters more than camera resolution
- Shoot straight-on, not at an angle
- For handwriting, ensure letters and numbers are clearly separated
The AI reads and interprets the problem
This step is where most of the work happens. The AI reads the full content of the image — text, mathematical notation, diagrams, labeled variables, and figures — and identifies what the problem is actually asking. This isn’t optical character recognition alone; the model understands context.
A projectile motion problem with a diagram gets read as a physics kinematics problem, not just a collection of numbers. A circuit schematic with labeled resistors gets read in the context of Ohm’s law and Kirchhoff’s rules. The subject tag you select (Math, Physics, Chemistry, etc.) helps the model apply the right framework from the start.
This interpretation step is what allows the tool to handle problems that can’t be typed — force diagrams, chemical structure drawings, geometry figures with labeled angles — because the visual context is part of the problem itself.
The solution is built step by step
Once the problem is understood, the AI works through it methodically — identifying the relevant principle or formula, stating what is known and unknown, substituting values, and performing calculations with units tracked throughout.
The output is structured to match how a good teacher or examiner would present the working: state the method, apply it, show each calculation step, verify if appropriate, state the final answer. This structure is consistent across subjects and problem types.
The response is deliberately verbose — not because we pad it, but because understanding requires seeing the reasoning at each stage, not just the transitions between numbers.
- Identification of the problem type and relevant method
- Statement of known variables and what is being solved for
- Formula or principle applied, with explanation of why it applies
- Step-by-step calculation with units at each stage
- Final answer clearly stated
You see the answer first, then the full walkthrough
The final answer is displayed immediately and clearly — so you know right away whether the problem was solved and what the result is. Below it, the complete step-by-step reasoning is available, showing exactly how the answer was reached.
This structure matters because it serves different needs. Sometimes you need to check whether your own answer is correct. Sometimes you need to understand where you went wrong in your working. The answer tells you the first; the steps tell you the second.
Reading through the steps is the part that actually builds understanding — not copying the answer, but following the method and recognising it for next time.
What makes the reading accurate
Visual understanding, not just OCR
Standard OCR reads characters. This goes further — it understands spatial relationships, like which number is a superscript vs. a coefficient, or whether a fraction bar separates a numerator from a denominator.
Handwriting recognition
The model is trained on handwritten academic content — equations, notes, diagrams. Typical classroom handwriting, even rushed or dense notation, is recognized accurately without needing a clean scan.
Diagram interpretation
Labeled diagrams — force arrows, circuit schematics, ray diagrams, coordinate geometry figures — are read in context. The labels and the geometry together inform the solution, not just the text.
Fast response time
From upload to solution is typically under 10 seconds for standard problems. Complex multi-step problems may take slightly longer, but the response is complete — not a streaming partial answer.
What it can solve
Dedicated pages for each subject with examples, tips, and subject-specific guidance.
Math Solver by Photo
Algebra, calculus, geometry, statistics, trigonometry, linear algebra — from Grade 8 through university.
Go to Math Solver →Physics Solver by Photo
Kinematics, mechanics, circuits, thermodynamics, optics, waves — including diagram and schematic reading.
Go to Physics Solver →Chemistry Solver by Photo
Balancing equations, stoichiometry, organic chemistry, thermochemistry, acid-base problems.
Try Chemistry Solver →Biology by Photo
Cell biology, genetics, Punnett squares, ecology, photosynthesis, cellular respiration.
Try Biology Solver →Homework Helper
All subjects in one place — history, English, languages, computer science, economics, and more.
Try Homework Helper →Word Problem Solver
Extract the math from written problems — rate, ratio, mixture, distance, profit, and applied problems.
Try Word Problem Solver →Questions about how it works
Ready to try it for yourself?
Upload a photo of any problem and see the full process in action.
📷 Try It Now — Free