Speak Math in Typst: Voice Input for the LaTeX Alternative | MathVoice

Speak Math in Typst: Voice Input for the LaTeX Alternative

Typst has won over thousands of writers with millisecond compiles and sane syntax. But if you dictate your math — or you simply hate typing backslashes — there's a deeper reason to like it: Typst's math syntax sounds like what you say. Here's why that matters, side-by-side examples, and a free way to dictate an equation and get Typst out.

By MathVoice · August 23, 2026 · 7 min read

1. Typst in 30 seconds

Typst is a new typesetting system positioned as the modern LaTeX alternative. Instead of minutes-long compile cycles, documents render as you type. Instead of \begin{document}, you just start writing. And instead of a fifty-year-old macro language, formulas look like this:

$ x = (-b plus.minus sqrt(b^2 - 4ac))/(2a) $

That's the quadratic formula. No \frac{}{}, no \pm, no \sqrt{} — just words arranged the way you'd say them. Since its 2023 public launch, Typst has accumulated a passionate community, package ecosystem, and growing adoption in university course notes, theses, and blogs.

But Typst has a gap that nobody has filled yet: input speed for math. Typing math in any system is slower than typing prose. Which raises an obvious question — why type it at all?

2. Why LaTeX math is hard to speak

Voice input for math has existed for a while, but almost all of it targets LaTeX. That's a problem, because spoken math and LaTeX syntax live on opposite sides of a translation gulf. When you say:

"the integral from zero to one of x squared d x"

a voice-to-LaTeX system must produce \int_{0}^{1} x^{2}\,dx — four backslash commands, three brace pairs, and a spacing macro, all reconstructed perfectly from audio. Every one of those symbols is a chance for the speech recognizer or the transcription model to slip. Braces are unbalanced, backslashes get dropped, and suddenly your equation doesn't compile.

Typist users know this pain from typing; voice users feel it even more because they never see the intermediate syntax. The fewer structural tokens a target format needs, the higher the end-to-end accuracy. Which brings us to the core argument:

Typst's math syntax is a phonetic near-match for mathematical speech. "Square root of b squared minus four a c" is sqrt(b^2 - 4ac) word for word. "The fraction a over b" is a/b. "Plus or minus" is literally plus.minus. The distance between what you say and what you write collapses to nearly zero.

3. Spoken math → LaTeX → Typst, side by side

You sayVoice → LaTeXVoice → Typst
"x squared plus one"x^{2} + 1x^2 + 1
"a over b"\frac{a}{b}a/b
"square root of b squared minus four a c"\sqrt{b^{2}-4ac}sqrt(b^2 - 4ac)
"integral from zero to one of x squared dx"\int_{0}^{1} x^{2}\,dxintegral_0^1 x^2 dif x
"sum from i equals one to n of one over i squared"\sum_{i=1}^{n} \frac{1}{i^{2}}sum_(i=1)^n 1/(i^2)
"limit as x approaches zero of sine x over x"\lim_{x \to 0} \frac{\sin x}{x}lim_(x -> 0) sin(x)/x
"plus or minus"\pmplus.minus
"alpha times beta"\alpha \cdot \betaalpha dot.op beta

Count the structural tokens in each column. In the limit example, the LaTeX form needs eleven characters of pure syntax — braces, backslashes, underscores — before any mathematics appears. The Typst form needs four parentheses' worth, and every token is a real English word. For a speech pipeline, that difference compounds: shorter outputs mean fewer transcription errors and cheaper, faster LLM calls.

4. Try it: dictate an equation, copy Typst

You can do this today, free, in MathVoice Studio:

  1. Open Studio in Chrome or Edge.
  2. Click New equation and dictate — "the sum from i equals one to n of one over i squared".
  3. In the Source panel on the right, click the new Typst tab.
  4. Hit Copy Typst and paste into any .typ file or directly into the typst.app editor inside $ … $.

The Chrome extension's sidebar gained the same superpower: after editing an equation on Overleaf, Notion, or any supported site, hit Copy as Typst and the converted source lands on your clipboard.

How it works: dictation still runs through MathVoice's LaTeX-native edit engine — the same role-based engine that powers surgical edits like "move minus b outside the fraction". Your dictated equation is stored internally as structured math; the Typst tab is a lossless-for-common-cases projection of that structure into Typst syntax. You get LaTeX-grade edit power with Typst-grade output.

5. A first for the Typst ecosystem

We searched for an existing speech-to-Typst tool and found nothing: voice math tools target LaTeX, and Typst tooling focuses on packages, templates, and converters. As far as we can tell this is the first time you can speak an equation into existence and get Typst out — which feels appropriate, given that Typst's own design philosophy is "make the common case sound natural."

If you're a Typst user, we'd genuinely like your feedback on the conversion quality — especially around matrices (mat(...)) and cases. If something converts wrong, tell us; the converter is a single well-tested module and iterates fast.

6. FAQ

Can I use voice input with Typst?

Yes — that's exactly what the Typst tab does. Dictate in Studio, open the Typst tab, copy the source.

Does MathVoice stop supporting LaTeX?

No. LaTeX remains the native format of the edit engine. Typst joins LaTeX and MathML as output formats; nothing about existing LaTeX workflows changes.

What about editing existing Typst documents?

Round-tripping arbitrary Typst files through the edit engine is on the roadmap (best-effort conversion exists today via our API, but complex documents may lose structure). Editing dictated-and-exported equations works fully, since MathVoice retains the structured original.

Is it free?

Yes — Studio and the Typst output are free at mathvoice.app/studio.

Dictate your next equation

Speak math, watch it render, copy Typst — free, no signup. Works best in Chrome or Edge.

Open MathVoice Studio →