A LaTeX engine built for handwritten math.
Most OCR tools read text linearly.
Mathematics is not linear.
Axiom’s LaTeX Engine is designed to interpret structure, not just characters—so handwritten equations become valid, compile-ready LaTeX, not approximations.
Why generic OCR fails for math
Traditional OCR systems are optimized for prose. They assume text flows left-to-right in a single hierarchy.
Mathematics violates these assumptions.
Common failures include:
- Fractions flattened into inline text
- Superscripts misread as adjacent characters
- Matrices collapsed into unreadable sequences
- Loss of alignment in multi-line equations
- Greek symbols misidentified or dropped
The result is output that looks digital but cannot compile.
What makes the LaTeX Engine different
Structure-aware recognition
The Axiom LaTeX Engine analyzes handwritten pages as spatial systems, not text streams.
It identifies:
- Baselines and alignment
- Vertical relationships (superscripts, subscripts, limits)
- Grouping boundaries (fractions, radicals, matrices)
- Symbol roles based on context, not shape alone
This allows the engine to reconstruct the logical hierarchy of an equation before generating LaTeX.
What the engine handles reliably
Supported mathematical structures
The LaTeX Engine is optimized for real academic notation, including:
- Fractions and nested fractions
- Superscripts and subscripts
- Definite and indefinite integrals
- Summations and limits
- Matrices and vectors
- Multi-line aligned equations
- Greek symbols and operators
- Physics-style notation and variables
Each structure is translated into standard LaTeX syntax, not a visual approximation.
From handwriting to compile-ready code
What “compile-ready” means
The output produced by Axiom is:
- Valid LaTeX source code
- Immediately compilable without cleanup
- Compatible with Overleaf and local LaTeX toolchains
- Free of proprietary wrappers or extensions
You can copy the output directly into:
- Overleaf
- TeX editors
- Markdown files with MathJax
- Version-controlled repositories
No retyping. No manual fixes.
What changes after processing
Before
A photographed page of handwritten equations.
After
A clean LaTeX document where:
- Fractions are preserved
- Alignment matches the original
- Symbols retain their meaning
- The document compiles successfully
The goal is not to “make it look nice.”
The goal is to make it correct.
Integration in real workflows
Designed to fit existing tools
The LaTeX Engine does not replace your workflow.
It removes the most fragile step in it.
Common uses include:
- Drafting equations by hand, then importing into Overleaf
- Converting exam notes into searchable study material
- Transcribing lab notebooks into formal documentation
- Moving handwritten derivations into research papers
The output remains plain text, so it works wherever LaTeX works.
Accuracy philosophy
Why accuracy depends on structure
Math recognition accuracy is not about sharper images or bigger models.
It depends on whether the system understands:
- That a symbol’s meaning changes with position
- That spacing and alignment carry semantic weight
- That notation is hierarchical, not sequential
The LaTeX Engine is trained specifically on STEM notation, not general handwriting.
What the LaTeX Engine is
- ✓A structural parser for handwritten math
- ✓A bridge between handwritten work and digital publication
- ✓A tool built for correctness, not demos
What it is not
- ×A generic OCR scanner
- ×A visual transcription tool
- ×A proprietary format generator
Use the LaTeX Engine
If you write equations by hand and need them in LaTeX,
Axiom eliminates the retyping step entirely.