If you've ever taken a graduate-level math, physics, or engineering course, you know the pain. The professor puts up a slide with a beautiful Hamiltonian or a triple integral with a Jacobian, and your first thought is: I'm going to need this in my thesis.
Your second thought: I'm going to have to retype this entire thing in LaTeX, aren't I.
And then you spend 20 minutes on it. And then you get a compile error. And then you spend 10 more minutes figuring out you had an extra closing brace somewhere in the middle of the subscript.
I did this for two years during my master's degree before I found a better way.
The Problem With the Existing Workflow
There are actually good tools for converting images of equations to LaTeX. Mathpix Snip is the most well-known, and the accuracy is genuinely impressive. But here's what the workflow actually looks like in practice:
- Take a screenshot of the slide
- Open the Mathpix app (or switch to the Mathpix tab)
- Paste the screenshot
- Wait for the conversion
- Copy the LaTeX output
- Switch back to your editor
- Paste into your .tex file
- Run the compiler
- Fix errors if any
That's nine steps. For every single equation. During a lecture where you're also trying to understand the material and take notes.
I also tried pix2tex, which is an open-source Python-based OCR tool that runs locally. The math: it works reasonably well once you've got it set up. The problem: getting it set up on Apple Silicon took me an hour of dealing with PyTorch version conflicts. And even after that, the workflow is still "separate tool → copy → paste → switch back to editor."
Google Gemini can do it too if you paste an image and ask nicely. Also a context switch.
The core issue isn't any of these tools specifically. It's that the OCR lives outside the editor. You're constantly breaking your flow to go get the LaTeX and bring it back.
What TeX64 Does Differently
TeX64 is a local-first LaTeX editor for macOS (https://tex64.com). I started using it primarily because I was tired of Overleaf's free-tier compile limits during my thesis crunch, but I quickly discovered it had built-in equation OCR.
The workflow is:
- Take a screenshot (
Cmd+Shift+4on macOS) - Paste it into TeX64 (
Cmd+V) - LaTeX code appears
That's it. Three steps. You never leave the editor.
Real Examples From My Coursework
Example 1: Partition Function (Statistical Mechanics)
The slide showed: Z = Σ_n exp(-β E_n)
TeX64 OCR output:
Z = \sum_{n} e^{-\beta E_n}
Exact. Ready to paste.
Example 2: Fourier Transform Definition
\hat{f}(\xi) = \int_{-\infty}^{\infty} f(x) \, e^{-2\pi i x \xi} \, dx
Including the \, spacing before dx. That's a nice detail I would have forgotten.
Example 3: Quantum State Vector Expansion
|\psi\rangle = \sum_{n=0}^{\infty} c_n |n\rangle, \quad c_n = \langle n | \psi \rangle
The bra-ket notation came out correctly — \langle, \rangle, the whole thing.
When OCR Gets It Wrong
It's not 100% perfect. Complex multi-line expressions sometimes have minor issues.
This is where TeX64's AI assistant, Axiom, becomes useful. When I get a compile error after an OCR paste, I can ask Axiom to look at the log. It reads the actual compile output and shows me a diff: "line 47, change this to that." One click to apply.
Everything — OCR, editing, error fixing — stays in one app.
Honest Limitations
- macOS only. Apple Silicon and Intel Mac only.
- You still need MacTeX or TeX Live. TeX64 is the editor, not the TeX distribution.
- OCR is for math, not text. It won't transcribe surrounding prose from a slide.
- Handwriting accuracy drops a bit. Clean PDF slides work much better than photographed handwriting.
- Free tier has limits. Heavy OCR usage requires a paid plan.
Why This Actually Matters
The difference between "screenshot → paste → done" and "screenshot → open app → paste → copy → switch app → paste → compile" is not just time. The first stays in flow. The second breaks it.
If you're writing LaTeX on a Mac and frequently pulling equations from slides or PDFs, it's worth trying TeX64 just for this feature.
Top comments (0)