| name | typst-slide |
| description | Typst slide workflow using Touying: create, compile, review, and polish academic presentations. Use this skill whenever the user works on Typst (.typ) slide decks, or asks to create slides, make a presentation, prepare a lecture, build a talk, or generate presentation materials using Typst or Touying. Also trigger when the user mentions Typst slides, Touying themes, CeTZ diagrams for slides, or wants to convert Beamer slides to Typst.
|
Typst Slide Workflow (Touying)
Universal skill for academic Typst presentations via Touying. Full lifecycle:
create โ compile โ review โ polish โ verify.
Why Typst over Beamer? Typst compiles in milliseconds (no multi-pass), has sane error messages, built-in scripting, and automatic package management. Touying is the most mature Typst slide framework with rich theme support.
0. REFERENCE PREAMBLE
Default template for new presentations. Use this unless the user provides a custom template.
#import "@preview/touying:0.7.0": *
#import themes.university: *
#import "@preview/cetz:0.4.2"
#import "@preview/fletcher:0.5.8" as fletcher: node, edge
#import "@preview/theorion:0.5.0": *
#import cosmos.clouds: *
#show: show-theorion
// CeTZ and Fletcher bindings for Touying animations
// Inside CeTZ canvas, use (pause,) tuple syntax โ NOT #pause
#let cetz-canvas = touying-reducer.with(reduce: cetz.canvas, cover: cetz.draw.hide.with(bounds: true))
// Inside Fletcher diagram, use bare pause (no #)
#let fletcher-diagram = touying-reducer.with(reduce: fletcher.diagram, cover: fletcher.hide)
// Semantic colors โ use throughout for visual consistency
#let positive = rgb("#0173B2") // blue (correct, advantage)
#let negative = rgb("#DE8F05") // orange (limitation, drawback)
#let emphasis = rgb("#029E73") // green (highlight, key finding)
#let neutral = luma(140) // muted context
#let accent = rgb("#CC78BC") // additional accent (purple)
// Semantic text helpers
#let pos(body) = text(fill: positive, body)
#let con(body) = text(fill: negative, body)
#let HL(body) = text(fill: emphasis, body)
#show: university-theme.with(
aspect-ratio: "16-9",
// handout: true โ uncomment for PDF export (collapses #pause subslides)
config-common(
handout: true,
frozen-counters: (theorem-counter,),
),
config-info(
title: [Title],
subtitle: [Subtitle],
author: [Presenter: \[name\]],
date: datetime.today(),
institution: [Shanghai Jiao Tong University],
logo: emoji.school,
),
)
#title-slide()
Rules:
1. HARD RULES (Non-Negotiable)
- Use
#pause sparingly; always compile with handout: true โ Touying's #pause creates animation subslides. In live presentation tools (pdfpc), these work as progressive reveals. But in normal PDF export, each subslide becomes a separate page โ a slide with 3 #pause produces 3 PDF pages, with the first showing only partial content. This is the #1 cause of "empty-looking" slides and inflated page counts. Always use config-common(handout: true) by default to collapse subslides. Only set handout: false when the user explicitly wants animation pages. For complex multi-step constructions, prefer separate slides over #pause. Inside CeTZ canvases, use (pause,) tuple syntax. Inside Fletcher diagrams, use bare pause. For list-item-by-item reveals, use #item-by-item[...].
- Max 2 colored/styled blocks per slide โ more dilutes emphasis. Demote transitional remarks to plain italic.
- Motivation before formalism โ every concept starts with "Why?" before "What?".
- Worked example within 2 slides of every definition.
- Typst source is the single source of truth โ CeTZ diagrams, content, notation all originate here.
- Verify after every task โ compile, check warnings, review PDF.
- Telegraphic style โ keyword phrases, not full sentences. Slides are speaker prompts, not manuscripts. Exception: framing sentences that set up a definition or transition. Each bullet item should be โค2 lines (~15 words) โ if longer, split into sub-items or rewrite more concisely.
- Every slide earns its place โ each slide must contain at least one substantive element (formula, diagram, table, theorem, or algorithm). A slide with only 3 short bullets and nothing else must be merged or enriched.
- Content density guard โ Typst's layout engine is good but not infallible. For slides mixing text and diagrams, always compile and visually verify. Common overflow happens with large CeTZ canvases below long equation blocks.
- Reference slide โ the second-to-last slide (before Thank You) must be a References slide listing key cited works. Use
#bibliography("refs.bib") or manual list with #text(size: 0.8em)[...].
- Color and contrast standards โ text-background contrast ratio โฅ 4.5:1 (WCAG AA). Never use red+green for binary contrasts. Prefer blue+orange. Semantic color helpers defined in preamble:
pos() = positive (blue), con() = negative (orange), HL() = emphasis (green). These are color-blind safe. Limit total palette to 3-5 colors.
- Backup slides โ after the Thank You slide, include 3-5 backup slides for anticipated questions. Use
#show: appendix before the backup section to freeze slide numbering. Content to include:
- Full proof of the main theorem (if only a sketch was shown)
- Parameter choices / security analysis details
- Extended comparison table with additional baselines
- Experimental setup details
- Definitions of prerequisites that were assumed known
2. ACTIONS
Parse <user request> to determine which action to run. If no action specified, ask.
2.1 compile [file]
Typst compiles in a single pass โ no multi-pass needed.
typst compile FILE.typ
For live preview during development:
typst watch FILE.typ
Post-compile checks:
- Check exit code (0 = success)
- Read any error/warning output from stderr
- Open PDF for visual verification
- Report: success/failure, any warnings, page count
If typst is not installed, guide the user:
brew install typst
cargo install typst-cli
2.2 create [topic]
Collaborative, iterative presentation creation. Strict phase gates โ never skip ahead.
Phase 0: Material Analysis (if papers/materials provided)
Read first, ask later. Must understand the content before asking meaningful questions.
- Read the full paper/materials thoroughly
- Extract: core contribution, key techniques, main theorems, comparison with prior work
- Map notation conventions
- Identify the paper's logical structure and which parts are slide-worthy
- Internally note: prerequisite knowledge, natural section boundaries, what could be skipped or expanded
Do NOT present results or ask questions yet โ proceed directly to Phase 1.
Phase 1: Needs Interview (MANDATORY โ informed by Phase 0)
Conduct a content-driven interview via AskUserQuestion. The questions below are the minimum required set โ you MUST also add paper-specific questions derived from Phase 0.
Minimum required questions (always ask):
- Duration: How long is the presentation?
- Audience level: Who are the listeners?
Optional question (ask when the talk is a journal club, defense, or user seems to want rehearsal support):
3. Speaker notes: Would you like speaker notes? If yes, #speaker-note[...] blocks will be added per slide with telegraphic talking points. Requires config-common(show-notes-on-second-screen: right).
Content-driven questions (derive from Phase 0, ask as many as needed):
- Prerequisite knowledge: List concrete technical dependencies. Ask which ones the audience knows.
- Content scope: Offer the paper's actual components as options. Ask which to emphasize, skip, or briefly mention.
- Depth vs. breadth: If the paper has both intuitive overview and detailed constructions, ask which the user prefers.
- Paper-specific decisions: E.g., if the paper compares two constructions, ask whether to present both equally or focus on one.
Guidelines:
- 3-6 questions total; don't over-ask, don't under-ask.
- If something is obvious from context (e.g., user said "่ฎจ่ฎบ็ญ"), infer rather than ask.
Slide count heuristic: ~1 slide per 1.5-2 minutes.
Timing allocation table:
| Duration | Total slides | Intro/Motivation | Methods/Background | Core content | Summary/Conclusion |
|---|
| 5min (lightning) | 5-7 | 1-2 | 0-1 | 2-3 | 1 |
| 10min (short) | 8-12 | 2 | 1-2 | 4-5 | 1 |
| 15min (conference) | 10-15 | 2-3 | 2-3 | 5-7 | 1-2 |
| 20min (seminar) | 13-18 | 3 | 2-3 | 6-9 | 2 |
| 45min (keynote) | 22-30 | 4-5 | 5-7 | 10-14 | 2-3 |
| 90min (lecture) | 45-60 | 5-6 | 8-12 | 25-35 | 3-4 |
Phase 2: Structure Plan (GATE โ user must approve before drafting)
Produce a detailed outline. For each section:
- Section title
- Number of slides allocated
- Key content points per slide (1-2 lines each)
- CeTZ/Fletcher diagrams planned (brief description)
- Notation to introduce
Present the plan to the user. Ask: structure OK? Expand/shrink/cut anything?
Do NOT proceed to drafting until user approves.
Phase 3: Draft (iterative, batched)
3a. Typst Slide Syntax
Heading-based slides (simple, recommended for most content):
= Section Title
== Slide Title
Content goes here. Use `#pause` for sequential reveals.
- First point
#pause
- Second point (appears on next click)
Explicit #slide syntax (for advanced layouts like multi-column):
#slide(composer: (1fr, 1fr))[
Left column content.
][
Right column content.
]
Key syntax differences from Beamer:
$x^2$ for inline math (same as Beamer's $x^2$)$ x^2 + y^2 = z^2 $ for display math (spaces around content trigger display mode)*bold* instead of \textbf{}_italic_ instead of \textit{}#text(fill: red)[colored text] instead of \textcolor{red}{}- Lists use
- (unordered) or + (ordered), with indentation for nesting
#table(columns: 3, ..cells) instead of \begin{tabular}#image("path.png", width: 80%) instead of \includegraphics
3b. Writing Style
- Telegraphic keywords, not full sentences. Exception: one framing sentence per slide to set context.
- Formulas and analysis interleave tightly โ define a quantity, then immediately state its cost/property/implication on the same slide.
- No conversational hedging โ never write "wait, not exactly" or "actually, let me clarify".
- Use
*bold* for key terms on first introduction; use #pos[...] for positive properties, #con[...] for drawbacks, #HL[...] for key findings.
3b-2. Opening and Closing Strategies
Opening slide (pick one strategy):
- Surprising statistic โ a counter-intuitive number that challenges assumptions
- Provocative question โ something the audience cannot immediately answer
- Real-world failure/problem โ "System X failed because..." โ "How do we prevent this?"
- Visual demonstration โ show the phenomenon before explaining it
Closing strategies:
- Call-back to opening โ revisit the opening question/problem, now answered
- 3 key takeaways โ numbered, telegraphic, one slide
- Open question / future direction โ leave the audience thinking
- Never end on a bare "Thank You" โ the second-to-last content slide should deliver the lasting impression
3c. Mathematical Slide Patterns
Definition slide:
== Definition: Polynomial Commitment
_Why_: We need a way to commit to a polynomial without revealing it.
#definition(title: "Polynomial Commitment")[
A polynomial commitment scheme is a tuple $(sans("Setup"), sans("Commit"), sans("Open"), sans("Verify"))$ ...
]
Key properties:
- *Binding*: cannot open to a different polynomial
- *Hiding*: commitment reveals nothing about $f$
Theorem/Proof slide:
== Main Result
_Informal_: Our scheme achieves linear prover time with logarithmic verification.
#theorem(title: "Main Theorem")[
For any polynomial $f$ of degree $n$, the prover runs in $O(n)$ field operations
and the verifier in $O(log n)$.
]
// Proof on the NEXT slide โ never cram theorem + proof on one slide
Comparison slide:
== Comparison with Prior Work
#table(
columns: (1fr, auto, auto, auto),
align: (left, center, center, center),
table.header[*Scheme*][*Prover*][*Verifier*][*Proof size*],
table.hline(),
[Prior work], [$O(n log n)$], [$O(n)$], [$O(n)$],
[*Ours*], [#pos[$O(n)$]], [#pos[$O(log n)$]], [#pos[$O(log n)$]],
)
3d. Content Density Constraints
Upper bounds (per slide):
- โค 7 bullet points or items
- โค 2 displayed equations (more โ split)
- โค 5 new symbols introduced
- โค 2 styled blocks (theorem/definition/alert boxes)
Lower bounds (per slide):
- Each slide MUST contain at least one substantive element: a formula, diagram, table, theorem, or algorithm
- A slide with only โค 3 short text-only bullets is too sparse โ merge or enrich
- Pure text-only slides should be โค 30% of the total deck
3e. Batch Workflow
- Work in batches of 5-10 slides, following the approved structure
- After each batch, compile (
typst compile) to catch errors early
- After each batch, self-check: notation consistency, density constraints, motivation-before-formalism
- Continue to next batch only after current batch compiles cleanly
3f. Table Best Practices
Typst tables are native and powerful โ no external packages needed.
#table(
columns: (2fr, 1fr, 1fr, 1fr),
align: (left, center, center, center),
stroke: none,
table.header[*Method*][*Time*][*Space*][*Accuracy*],
table.hline(),
[Baseline], [$O(n^2)$], [$O(n)$], [72%],
[*Ours*], [#pos[$O(n)$]], [$O(n)$], [#pos[*89%*]],
table.hline(),
)
- Always use
stroke: none + table.hline() for academic-style tables (like booktabs)
- Column alignment: numbers centered, text left-aligned
- Max 6-7 columns, 8-10 rows per slide
- Highlight best results with
*bold* or #pos[...]
- For wide tables: use
#utils.fit-to-width(1fr)[#table(...)] to auto-scale
3g. Algorithm Display
#import "@preview/algorithmic:0.1.0"
#import algorithmic: algorithm
#algorithm({
import algorithmic: *
Function("Verify", args: ($pi$, $x$), {
Cmt[Parse the proof]
Assign[$c$][$pi.c$]
If(cond: $e(c, g) = e(pi.h, tau)$, {
Return[$sans("accept")$]
})
Return[$sans("reject")$]
})
})
- Pseudocode โค 10 lines per slide
- Highlight critical lines with color
- Input/output clearly stated
Phase 4: Figures
- CeTZ for geometric diagrams, coordinate plots, annotated figures
- Fletcher for flowcharts, state diagrams, node-edge graphs
- Apply CeTZ quality standards (see Section 2.6)
Data visualization guidelines (same as Beamer skill โ projector-optimized):
- Simplify ruthlessly โ remove minor gridlines, label directly on plot
- Enlarge everything โ labels, line widths, markers must be readable at distance
- One message per figure โ split multi-panel figures across slides
- Color-blind safe palette: blue (#0173B2) + orange (#DE8F05), with line-style differences as redundant encoding
- For data-driven plots: use CeTZ
cetz.plot or consider generating with Python and including as images
Phase 5: Quality Loop (MANDATORY โ iterative)
After completing the full draft, enter the quality loop:
โโโ 5a. Compile (typst compile)
โ 5b. Self-Review (structure + content + visual)
โ 5c. Score (apply rubric)
โ 5d. Fix all issues found
โโโ If score < 90 and round < 3: loop back to 5a
If score โฅ 90 or round = 3: report to user
5a. Compilation
typst compile FILE.typ โ single pass, check exit code and stderr- Open PDF for visual inspection
5b. Self-Review โ re-read the .typ and verify:
Structure:
Content density:
CeTZ and visuals:
Notation:
5c. Quality Score
Start at 100. Deduct:
| Severity | Issue | Deduction |
|---|
| Critical | Compilation failure | -100 |
| Critical | Content overflow (equation or diagram) | -20 |
| Critical | CeTZ diagram overflows slide boundary | -15 per diagram |
| Critical | Undefined variable / import error | -15 |
| Major | Content visually clipped in PDF | -10 per slide |
| Major | CeTZ computed points not on curve | -8 per diagram |
| Major | Sparse slide (โค3 items, no math/diagram) | -5 per slide |
| Major | CeTZ label overlap | -5 |
| Major | Missing references slide | -5 |
| Major | Notation inconsistency | -3 |
| Minor | Excessive spacing hacks | -1 |
| Minor | Font size reduction overuse | -1 per slide |
Thresholds:
- โฅ 90: Ready to deliver. Report to user.
- 80-89: Acceptable. Fix remaining majors if possible.
- < 80: Must fix. Loop back and resolve critical/major issues.
Post-Creation Checklist (final gate)
[ ] Compiles without errors
[ ] All imports resolve (packages downloaded)
[ ] Score โฅ 90
[ ] Every definition has motivation + worked example
[ ] Max 2 styled blocks per slide
[ ] No sparse slides
[ ] CeTZ diagrams visually verified โ no overlaps, no overflow
[ ] Tables fit within slide boundaries
[ ] References slide present (second-to-last, before Thank You)
2.3 review [file] or proofread [file]
Read-only report, no file edits.
5 check categories:
| Category | What to check |
|---|
| Grammar | Subject-verb, articles, prepositions, tense consistency |
| Typos | Misspellings, duplicated words, unreplaced placeholders ([name], [TODO], template remnants) |
| Overflow | Long equations, too many items per slide |
| Consistency | Citation format, notation, terminology, block usage |
| Academic quality | Informal abbreviations, missing words, claims without citations |
Report format per issue:
### Issue N: [Brief description]
- **Location:** [slide title or line number]
- **Current:** "[exact text]"
- **Proposed:** "[fix]"
- **Category / Severity:** [Category] / [High|Medium|Low]
2.4 audit [file]
Visual layout audit. Read-only report.
Check dimensions:
- Overflow: Content exceeding boundaries, wide tables/equations
- Font consistency: Inline size overrides, inconsistent sizes
- Block fatigue: 2+ styled blocks per slide
- Spacing: Manual spacing overuse, structural issues
- Layout: Missing transitions, missing framing sentences
2.5 pedagogy [file]
Holistic pedagogical review. Read-only report.
13 patterns to validate:
| # | Pattern | Red flag |
|---|
| 1 | Motivation before formalism | Definition without context |
| 2 | Incremental notation | 5+ new symbols on one slide |
| 3 | Worked example after definition | 2 consecutive definitions, no example |
| 4 | Progressive complexity | Advanced concept before prerequisite |
| 5 | Fragment reveals | Dense theorem revealed all at once |
| 6 | Standout slides at pivots | Abrupt topic jump, no transition |
| 7 | Two-slide strategy for dense theorems | Complex theorem crammed in 1 slide |
| 8 | Semantic color usage | Binary contrasts in same color |
| 9 | Block hierarchy | Wrong block type for content |
| 10 | Block fatigue | 3+ blocks on one slide |
| 11 | Socratic embedding | Zero questions in entire deck |
| 12 | Visual-first for complex concepts | Notation before visualization |
| 13 | Side-by-side for comparisons | Sequential slides for related definitions |
2.6 cetz [file]
CeTZ diagram review โ the Typst equivalent of the Beamer skill's TikZ review.
Quality standards:
- Labels NEVER overlap with curves, lines, dots, or other labels
- Visual semantics: solid=observed, dashed=counterfactual, filled=observed, hollow=counterfactual
- Line weights: use
stroke: 1.5pt for data, stroke: 1pt for axes, stroke: 0.5pt for grid
- Minimum spacing between labels and graphical elements
- Mathematical accuracy: NEVER hardcode y-coordinates for points on curves. Compute from the same function.
Common CeTZ patterns:
Coordinate plot with axes:
#cetz.canvas({
import cetz.draw: *
// Axes
line((-0.5, 0), (6, 0), mark: (end: "stealth"), stroke: 1pt)
line((0, -0.5), (0, 4), mark: (end: "stealth"), stroke: 1pt)
content((6, -0.3), $x$)
content((-0.3, 4), $y$)
// Curve
merge-path({
for x in range(1, 51) {
let xv = x / 10
let yv = calc.sqrt(xv)
line((xv, yv),)
}
}, stroke: positive + 1.5pt)
})
CeTZ with Touying animation (using touying-reducer):
#slide[
#cetz-canvas({
import cetz.draw: *
rect((0, 0), (5, 5))
(pause,) // <-- tuple syntax, NOT #pause
circle((2.5, 2.5), radius: 1)
(pause,)
line((0, 0), (5, 5), stroke: red)
})
]
Flowchart with Fletcher (preferred for node-edge diagrams):
#fletcher-diagram(
node-stroke: 0.5pt,
node-corner-radius: 3pt,
spacing: (15mm, 10mm),
node((0, 0), [Input $x$]),
node((1, 0), [Commit]),
node((2, 0), [Prove]),
node((3, 0), [Verify]),
edge((0, 0), (1, 0), "->", label: [encode]),
edge((1, 0), (2, 0), "->"),
edge((2, 0), (3, 0), "->", label: [$pi$]),
)
Annotated brace:
#cetz.canvas({
import cetz.draw: *
// ... content ...
// Use cetz.decorations for braces
line((1, 0), (1, 3), stroke: (dash: "dashed", paint: neutral))
content((1.3, 1.5), text(size: 0.8em)[annotation])
})
CeTZ checklist:
[ ] No label-label overlaps
[ ] No label-curve overlaps
[ ] Diagram fits within remaining slide space
[ ] ALL marked points computed from functions โ no hardcoded y-values
[ ] Consistent visual semantics (solid/dashed, filled/hollow)
[ ] Arrow annotations: FROM label TO feature
[ ] Axes extend beyond all data points
[ ] Labels legible at presentation size
CeTZ vs Fletcher decision guide:
- CeTZ: coordinate plots, geometric constructions, custom diagrams, function graphs
- Fletcher: flowcharts, state machines, protocol diagrams, any node-edge graph
- When in doubt, use Fletcher for node-edge and CeTZ for everything else
2.7 excellence [file]
Comprehensive multi-dimensional review. Use the Agent tool to dispatch review agents in parallel.
Parallel agent dispatch:
- Visual audit agent โ overflow, font consistency, block fatigue, spacing, transitions
- Pedagogical review agent โ 13 patterns, deck-level checks
- Proofreading agent โ grammar, typos, citation consistency, notation
- CeTZ review agent (if file contains CeTZ/Fletcher) โ label overlaps, geometric accuracy, spacing
- Domain review agent (optional) โ substantive correctness
After all agents return, synthesize a combined report with quality score.
2.8 visual-check [file]
PDFโimage systematic visual review. Converts compiled PDF to images, then reviews each slide visually. This catches issues invisible in source code โ Typst's layout engine rarely warns about visual overflow.
Why this matters: Unlike LaTeX, Typst doesn't emit overfull hbox warnings. Content that exceeds the slide boundary is silently clipped. The only way to catch these issues is visual inspection.
Workflow:
-
Compile (if not already compiled):
typst compile FILE.typ
-
Convert PDF to images using PyMuPDF:
import fitz
doc = fitz.open('FILE.pdf')
n = doc.page_count
zoom = 200 / 72
matrix = fitz.Matrix(zoom, zoom)
for i in range(n):
page = doc.load_page(i)
pixmap = page.get_pixmap(matrix=matrix)
pixmap.save(f'/tmp/slide-{i+1:03d}.jpg', output='jpeg')
doc.close()
print(f'Exported {n} slides')
Or via bash: python3 -c "import fitz; ..."
Fallback (if PyMuPDF unavailable): Use the Read tool directly on the PDF โ Claude Code is multimodal and can read PDF files page by page with the pages parameter (max 20 pages per request).
-
Systematic per-slide inspection โ use Read tool to view each image, checking:
-
Report per issue:
### Slide N: [slide title]
- **Issue:** [description]
- **Severity:** Critical / Major / Minor
- **Fix:** [specific recommendation]
Tip: For iterative editing, use typst watch FILE.typ to auto-recompile on save, then re-read the PDF periodically.
2.9 validate [file] [duration]
Automated quantitative validation.
-
Compile and count pages:
typst compile FILE.typ && python3 -c "import fitz; print(fitz.open('FILE.pdf').page_count)"
Compare against timing allocation table.
-
Source code static checks:
- Count
#pause usage โ flag if excessive (>3 per slide average)
- Check for missing bibliography/references section
- Count slides with >2 styled blocks
- Check all imports resolve
2.10 extract-figures [pdf] [pages]
Extract figures from a paper PDF for inclusion in Typst slides.
Workflow same as Beamer skill, but generate Typst snippets:
Full-width figure:
== Frame Title
#align(center)[
#image("figures/fig-label.png", width: 85%)
]
#text(size: 0.8em)[Source: Author et al. (2024)]
Figure + text side-by-side:
#slide(composer: (1fr, 1fr))[
#image("figures/fig-label.png", width: 100%)
#text(size: 0.8em)[Source: Author et al.]
][
Key observations:
- Point 1
- Point 2
]
3. VERIFICATION PROTOCOL
Every task ends with verification. Non-negotiable.
[ ] Compiled without errors (typst exit code 0)
[ ] No warnings in stderr
[ ] All imports resolve (packages auto-downloaded)
[ ] PDF opens and renders correctly
[ ] Visual spot-check of modified slides
4. DOMAIN REVIEW (Template)
Same 5 lenses as Beamer skill:
- Assumption stress test โ all assumptions stated, sufficient, necessary?
- Derivation verification โ each step follows?
- Citation fidelity โ slide accurately represents cited paper?
- Code-theory alignment โ code implements exact formula from slides?
- Backward logic check โ read conclusionโsetup, every claim supported?
5. TROUBLESHOOTING
Error: PDF has way too many pages / slides look half-empty
Cause: #pause creates animation subslides โ each becomes a separate PDF page. A slide with 4 pauses = 4 pages in the PDF, with the first 3 showing only partial content.
Fix: Add config-common(handout: true) to the theme .with() call. This collapses all subslides into one page with all content visible. Only disable for live presentation with pdfpc/pympress.
Error: Outline slide looks like a textbook table of contents (page numbers, dotted leaders, nested section numbers)
Cause: Using #outline() auto-generates a book-style TOC. Combined with numbly heading numbering, it produces "1.1 Title ......... 8" formatting.
Fix: Never use #outline() for slides. Write a manual bullet list instead. Never use numbly or #set heading(numbering: ...).
Error: #let myvar = $bb(G)$ then $#myvar_1$ produces error or wrong output
Cause: Top-level #let assigns a content value; using # inside math mode to call it doesn't compose well with subscripts/superscripts.
Fix: Write the math directly: $bb(G)_1$. Or define math-safe functions: #let GG(sub) = $bb(G)_#sub$ and call as #GG[1] outside math mode.
Error: unknown import or package not found
Cause: Wrong package version or name.
Fix: Check Typst Universe (https://typst.app/universe) for correct package name and latest version. Typst auto-downloads packages on first compile.
Error: content does not fit or visual overflow
Cause: Too much content for the slide.
Fix priority:
- Split into two slides
- Shorten text (telegraphic style)
- Use
#utils.fit-to-height(1fr)[...] or #utils.fit-to-width(1fr)[...] to auto-scale
- Reduce font size with
#text(size: 0.9em)[...] (last resort)
Error: Math rendering issues
Cause: Typst math syntax differs from LaTeX.
Common conversions:
\mathbb{F} โ bb(F)\mathcal{O} โ cal(O)\text{...} inside math โ "text" (quoted strings in math mode)\left( \right) โ lr(( )) (use lr() for auto-sizing delimiters)\sum_{i=0}^{n} โ sum_(i=0)^n\frac{a}{b} โ a / b or frac(a, b)\begin{align} โ use $ ... \ ... $ with \ for line breaks and & for alignment\binom{n}{k} โ binom(n, k)\langle, \rangle โ angle.l, angle.r\| \| โ norm(...) or lr(bar.double ... bar.double)
Error: CeTZ diagram not rendering
Cause: Missing import or wrong canvas syntax.
Fix: Ensure both #import "@preview/cetz:0.4.2" at top level AND import cetz.draw: * inside cetz.canvas({...}).
Error: #pause not working inside CeTZ canvas
Cause: CeTZ uses a different pause mechanism.
Fix: Use (pause,) tuple syntax inside CeTZ canvases (not #pause). For Fletcher diagrams, use bare pause.
Error: #pause not working inside context expression
Cause: Touying's #pause cannot penetrate context boundaries.
Fix: Use callback-style: #slide(repeat: N, self => [#let (uncover, only) = utils.methods(self); ...]). Must manually set repeat count.
Error: Page layout broken / set page(...) has no effect
Cause: Touying overrides set page().
Fix: Use config-page(...) inside the theme .with() call instead. E.g., config-page(margin: (x: 2em, y: 1em)).
Error: Theorems not numbered correctly with animations
Cause: Counter increments on each subslide.
Fix: Add config-common(frozen-counters: (theorem-counter,)) to theme configuration.
Error: Bibliography not rendering
Cause: Missing .bib file or wrong path.
Fix: Ensure the .bib file exists and use #bibliography("refs.bib") at the end. Typst supports BibTeX and Hayagriva formats. Alternative: use config-common(show-bibliography-as-footnote: bibliography("refs.bib")) to show citations as footnotes instead of a bibliography page โ then cite with @key in text.
Error: Touying slide not splitting on headings
Cause: Heading level mismatch with theme expectations.
Fix: Most Touying themes use = for sections and == for slides (slide-level=2). Exception: dewdrop uses slide-level=3 (= section, == subsection, === slide). For explicit control, use #slide[...].
Error: **bold** renders as colored alert text instead of normal bold
Cause: Some themes enable show-strong-with-alert: true by default.
Fix: Add config-common(show-strong-with-alert: false) to the theme .with() call.
Error: Figure/theorem numbers jump during animation
Cause: Counters increment on each subslide.
Fix: Freeze the relevant counter: config-common(frozen-counters: (theorem-counter,)) or config-common(frozen-counters: (figure.where(kind: image),)).
6. TYPST โ BEAMER MIGRATION CHEATSHEET
Quick reference for converting between the two systems:
| Beamer | Typst (Touying) |
|---|
\begin{frame}{Title} | == Title or #slide[...] |
\frametitle{T} | == T (heading-based) |
\pause | #pause |
\begin{itemize} | - item (dash lists) |
\begin{enumerate} | + item (plus lists) |
\textbf{bold} | *bold* |
\textit{italic} | _italic_ |
\texttt{code} | `code` |
\begin{block}{T} | #block(title: [T])[...] or theorem envs |
\begin{alertblock}{T} | #alert[...] or #block(fill: red.lighten(90%))[...] |
\begin{columns} | #slide(composer: (1fr, 1fr))[...][...] |
\includegraphics | #image("path", width: 80%) |
\begin{tabular} | #table(columns: ..., ...) |
\begin{tikzpicture} | #cetz.canvas({...}) |
\begin{align} | $ ... \ ... $ with & alignment |
\cite{key} | @key (with bibliography) |
\footnote{text} | #footnote[text] |
\hyperlink{label}{text} | #link(label("name"))[text] |
xelatex (3-pass) | typst compile (single pass, ~100ms) |
