python-pptx2 3.1.0__py3-none-any.whl → 3.2.0__py3-none-any.whl

pptx2/skill/references/slide-design.md

@@ -1,495 +1,550 @@
-# Slide design: making a generated deck look designed
-
-`space-aware-authoring.md` makes a deck *physically* correct — nothing
-overflows, nothing falls off the slide. This file is about the other
-half: a deck that looks like a designer laid it out. Everything here is
-built from the public API (`BBox`, `add_text`, `add_card`, `add_bullets`,
-`add_picture_fit`, `PALETTES`, `pptx2.diagrams`, `render_contact_sheet`),
-so the code samples run as written.
-
-Read this when the deck will be *seen* — a lesson, a talk, a pitch — and
-not only parsed.
-
-## The one-paragraph brief
-
-A slide that looks designed has **one idea**, **one focal point**, **one
-palette**, **one type family**, generous **air** around everything, and
-text that **fits at a readable size**. The viewer knows within a second
-where to look first. Every element on the slide is *content* (a title, a
-claim, a list, a picture, a table, a diagram, an equation) or *structure*
-(a surface, a rule, a caption); there is nothing whose only job is to
-decorate. When the content is strong the slide needs nothing else, and
-when the content is thin, ornament will not save it.
-
-## 1. Start from the content
-
-Before drawing anything, write down for each slide:
-
-- **Purpose** — one sentence: what the viewer should know or feel after it.
-- **Carrier** — the single element that delivers the purpose: a
-  statement, a short list, a picture, a table, a process, an equation.
-- **Support** — at most two more elements that help the carrier (a
-  caption, a source, a question, a picture beside a list).
-- **Archetype** — pick one from the catalog in §10. Reusing a handful of
-  archetypes is what makes a deck feel consistent.
-
-A deck of 8–12 slides typically needs 4–5 distinct archetypes. Alternate
-dense and light slides: after a table or a three-card slide, give the
-audience a statement or a full-bleed picture.
-
-## 2. Canvas, margins, grid
-
-Work in 16:9 at 13.333 × 7.5 in, and set it explicitly:
-
-```python
-from pptx2 import Presentation, BBox
-from pptx2.util import Inches
-
-prs = Presentation()
-prs.slide_width, prs.slide_height = Inches(13.333), Inches(7.5)
-W, H = 13.333, 7.5
-BLANK = prs.slide_layouts[6]
-```
-
-Then let two regions govern every content slide:
-
-```python
-TITLE = BBox.from_inches(0.8, 0.55, W - 1.6, 0.9)     # title band
-BODY  = BBox.from_inches(0.8, 1.7,  W - 1.6, H - 2.4)  # content area (ends at 6.8")
-```
-
-- **Margins**: 0.8 in left/right, 0.55 in top, 0.7 in bottom. Nothing but
-  a full-bleed picture or a coloured background touches the slide edge.
-- **Gutters**: 0.4–0.6 in between columns and cards. Use one value per deck.
-- **Columns**: `BODY.columns(3, gap=Inches(0.5))`,
-  `BODY.split_h([7, 5], gap=Inches(0.6))`, `BODY.rows(2, gap=...)`. Every
-  card, column and picture on a slide starts at the same left edge as the
-  title, or at a grid line derived from it.
-- **Optical center**: single statements and hero titles sit slightly
-  above the geometric middle (`anchor="middle"` in a box whose top is
-  ~1.6 in and bottom ~5.4 in).
-
-## 3. Typography
-
-One family for the whole deck. Noto Sans (or whatever the environment's
-default sans is) is a fine choice; the design comes from *scale* and
-*spacing*, not from the font.
-
-| Role | Size | Weight | Notes |
-|---|---|---|---|
-| Hero title (title / section slide) | 44–54 pt | bold | 1–2 lines |
-| Statement (the one-sentence slide) | 36–44 pt | bold | ≤ 20 words, centred or left |
-| Slide title | 30–36 pt | bold | one line; sentence case |
-| Body / bullets | 20–24 pt | regular | ≤ 6 items, ≤ 12 words each |
-| Card title | 20–24 pt | bold | |
-| Card body | 16–18 pt | regular | 2–4 lines |
-| Table cells | 16–20 pt | header bold | |
-| Caption / source / footer | 12–14 pt | regular or italic, `muted` | |
-
-Rules that make type look intentional:
-
-- **Sentence case** for titles ("Co rostlina potřebuje", not "CO ROSTLINA
-  POTŘEBUJE"). Uppercase belongs to acronyms.
-- **Left-align running text.** Centre only short statements, titles on
-  hero slides, and labels inside diagram nodes.
-- **Breathing room**: `line_spacing = 1.1–1.2` for body, `space_after =
-  8–10 pt` between bullets (`add_bullets(gap_pt=8)` does this).
-- **Zero text-box margins** when the box is already placed by the grid
-  (`margin_pt=0`) so the text edge *is* the grid line.
-- **Hierarchy by size and weight, not by colour.** Reserve colour for the
-  one thing on the slide that should pop (§4).
-- **Fit, then check**: `add_bullets` and `add_card` shrink text to the box
-  and never below 12 pt. If they had to shrink, there is too much text —
-  split the slide.
-
-```python
-slide.shapes.add_text(TITLE, text="Co rostlina potřebuje",
-                      size_pt=32, bold=True, color=P.ink,
-                      anchor="middle", margin_pt=0)
-```
-
-## 4. Colour
-
-Pick **one** palette and use only its roles:
-
-```python
-from pptx2 import PALETTES
-
-P = PALETTES["slate"]      # slate · linen · forest · plum · ember · navy · graphite
-D = P.dark()               # same hues on a dark paper — title, section, closing slides
-```
-
-| Role | Use |
-|---|---|
-| `P.paper` | slide background (`slide.background.fill.solid(); ...fore_color.rgb = P.paper`) |
-| `P.ink` | all headings and body text |
-| `P.muted` | captions, sources, secondary labels, table grid text |
-| `P.surface` | card fills, table zebra rows, picture frames |
-| `P.line` | hairline rules and table borders |
-| `P.accent` | **one** emphasis per slide: the key number, the active step, a highlighted word, the header row |
-| `P.accent_soft` | fill of a callout card / highlighted row; text on it stays `P.ink` |
-| `P.accent_ink` | text placed *on* `P.accent` (white on a dark accent) |
-
-- **One accent per slide.** Two accents split attention; three look like a
-  toy. Cards in a row all share `P.surface`; the one to emphasise (if any)
-  gets `P.accent_soft`.
-- **Dark slides** for the title, section dividers and the closing slide
-  give the deck rhythm. Use `D = P.dark()` so the hues stay in the family.
-- **Contrast**: every text/background pair ≥ 4.5:1 (the linter reports
-  `LowContrast`); the shipped palettes satisfy this for `ink` on `paper`,
-  `surface` and `accent_soft`, for `muted` on `paper`, and for `accent_ink`
-  on `accent`.
-- **Photos bring their own colour.** Beside a photo, keep the rest of the
-  slide to `paper` / `ink` / `muted`.
-
-## 5. Surfaces: cards, callouts, panels
-
-A card is one tinted rounded rectangle with padded text inside it, and
-that is the whole card. Its silhouette stays a clean rectangle from every
-angle; emphasis comes from the tint (or, when the tint is unwanted, a
-hairline outline — one or the other, never both). Padding of 18–28 pt on
-every side gives the text room to sit rather than press against the edge.
-
-```python
-from pptx2 import add_card
-
-for cell, (title, body) in zip(BODY.columns(3, gap=Inches(0.5)), items):
-    add_card(slide, cell, title=title, body=body,
-             fill=P.surface, title_color=P.ink, body_color=P.ink,
-             title_size_pt=22, body_size_pt=18, pad_pt=22)
-```
-
-Composition rules for groups of cards:
-
-- **2–4 cards per row**, equal size, equal gap, same fill, same radius.
-  Five or more become a table.
-- **Same content shape in every card**: if one has a title and two lines,
-  all have a title and two lines. Trim copy to match.
-- **A callout** (question, key definition, "remember this") is one wide
-  card in `P.accent_soft`, inset from the body region so it reads as a
-  deliberate object, not as a background.
-- **Cards hold content, backgrounds hold cards.** A card sits directly on
-  the paper; it does not sit on another panel.
-- **Numbering** belongs *in* the title text ("1 · Zachycení"), where it
-  aligns with everything else, rather than in a separate badge shape.
-
-`add_card` returns `card.inner` (the padded box) so a picture, equation
-or diagram can be placed inside the same surface:
-
-```python
-c = add_card(slide, right_col, fill=P.surface)
-slide.shapes.add_equation(c.inner, latex=r"6\,CO_2 + 6\,H_2O \rightarrow C_6H_{12}O_6 + 6\,O_2", size_pt=28)
-```
-
-## 6. Whitespace and alignment
-
-- 30–40 % of a content slide is empty. That is the design working.
-- **One left edge.** Title, bullets, cards and pictures share the `0.8 in`
-  margin (or a grid line derived from it). Right edges align to the
-  mirror margin.
-- **Consistent vertical rhythm.** Title band → 0.25 in gap → body. Every
-  content slide places its body at the same `BODY.top`, so flipping
-  through the deck feels steady.
-- **Balance, not symmetry.** A picture on the right with a list on the
-  left is balanced when the list's text block is roughly as tall as the
-  picture. Use `split_h([7, 5])` for text-heavy pairs, `[1, 1]` for a
-  true comparison.
-- **Empty is fine, unbalanced is not.** A slide whose right half is blank
-  because the list is short: make the list larger (24 pt), add a picture,
-  or narrow the text column and centre the pair.
-
-## 7. Pictures
-
-Photos carry emotion and specificity — one good photo does more than three
-small ones.
-
-```python
-from pptx2 import add_picture_fit
-
-# Beside text: fit inside the column, never distort, caption in muted.
-add_picture_fit(slide, "leaf.jpg", right_col, mode="contain",
-                caption="Buk lesní · Wikimedia Commons", caption_color=P.muted)
-
-# Full-bleed hero: crop to fill the whole slide, then a dark band for text.
-with slide.design_group("hero"):             # photo + band + title are one deliberate stack
-    add_picture_fit(slide, "forest.jpg", BBox.from_inches(0, 0, W, H), mode="cover")
-    band = slide.shapes.add_shape(MSO_SHAPE.RECTANGLE, *BBox.from_inches(0, H - 2.2, W, 2.2))
-    band.fill_hex(D.paper); band.line.fill.background(); band.shadow.clear()
-    band.fill.fore_color.alpha = 0.8         # translucent, photo shows through
-    slide.shapes.add_text(BBox.from_inches(0.8, H - 1.9, W - 1.6, 1.6),
-                          text="Les jako plíce planety", size_pt=44, bold=True,
-                          color=D.ink, anchor="middle", margin_pt=0)
-```
-
-- **`mode="contain"`** when the whole picture matters (diagrams, maps,
-  artworks); **`mode="cover"`** for atmosphere (backgrounds, hero slides,
-  square tiles in a grid).
-- **Picture grids**: `BODY.columns(3)` → three `mode="cover"` tiles of
-  equal size, one caption each.
-- **Caption every photo** that is not purely decorative: what it shows and
-  the source, 12–14 pt, `P.muted`.
-- Leave pictures unframed. A `P.surface` background behind a letter-boxed
-  `contain` picture is the one exception, so the empty band reads as a
-  frame rather than a gap.
-
-## 8. Tables, charts, diagrams
-
-**Tables** — structure carries the styling:
-
-```python
-tbl = slide.shapes.add_table(rows, cols, *BODY.inset(bottom=Inches(1.0)), style="clean").table
-tbl.format_cells(rows=0, fill=P.ink, color=P.paper, bold=True, size_pt=18)
-tbl.format_cells(rows=slice(1, None), size_pt=18, color=P.ink)
-tbl.format_cells(rows=slice(1, None, 2), fill=P.surface)      # zebra
-tbl.format_cells(rows=slice(1, None), cols=0, bold=True)      # row labels
-```
-
-≤ 6 body rows and ≤ 4 columns on a slide; more belongs in a handout. The
-header row is the accent of the slide — use `P.ink` or `P.accent`, not
-both.
-
-**Diagrams** — `pptx2.diagrams` recipes already space nodes and route
-arrows; feed them the palette:
-
-```python
-from pptx2.diagrams import horizontal_pipeline, cycle, hub_and_spoke, comparison_columns
-
-horizontal_pipeline(slide, BODY.inset(top=Inches(1.4), bottom=Inches(1.4)),
-                    steps=["Vstupy", "Světlo + chlorofyl", "Výstupy"],
-                    accent=P.accent, fill=P.surface, text_color=P.ink,
-                    card_line=None, card_radius=10, size_pt=20)
-```
-
-A pipeline with 3–5 steps, a cycle with 3–6 nodes, a hub with 3–6 spokes.
-Node labels are 2–4 words; explanations go in a list or a card beside the
-diagram.
-
-**Charts** — `chart.recolour([P.accent, P.muted, P.line])`, gridlines in
-`P.line`, one series in `P.accent` and the rest in greys when there is a
-story to tell. Title the chart with the takeaway ("Kyslík roste s
-osvětlením"), not the variable names.
-
-**Equations** — `slide.shapes.add_equation(bb, latex=..., size_pt=28)` or
-`paragraph.add_math(...)`; display equations at 28–36 pt, centred, with
-air around them.
-
-## 9. Overlaps: deliberate and accidental
-
-Overlap is a normal design move — text on a card, a caption on a photo, a
-translucent band over a hero image, a number disc on a timeline. The
-linter's `ShapeCollision` warning asks one question: *did you mean it?*
-
-- **Meant it** → say so, and the warning disappears:
-
-  ```python
-  with slide.design_group("hero"):           # everything created inside is one cluster
-      add_picture_fit(slide, "forest.jpg", BBox.from_slide(slide), mode="cover")
-      band = slide.shapes.add_shape(MSO_SHAPE.RECTANGLE, *BBox.from_inches(0, H - 2.2, W, 2.2))
-      slide.shapes.add_text(BBox.from_inches(0.8, H - 1.9, W - 1.6, 1.6), text="…")
-
-  slide.lint_group_overlaps(card, picture, caption)   # shapes already made, widest first
-  badge.allow_overlap_with(card)                      # exactly this pair
-  card.layer, badge.layer_above = "card", "card"      # also asserts z-order
-  ```
-
-  `add_card`, `add_picture_fit`, the components and the diagram recipes
-  already tag their own shapes, so a card and its text never trigger it.
-
-- **Did not mean it** → it is a layout bug: two text boxes crossing, a
-  card poking into its neighbour, a picture over a title. Fix the
-  geometry (use `columns()` / `split_h()` instead of hand arithmetic) and
-  re-lint.
-
-Deliberate stacking looks deliberate when the upper shape is clearly
-*inside* the lower one (inset on every side) or clearly *anchored* to one
-of its corners/edges. A shape that half-crosses a boundary reads as a
-mistake even when it is intentional.
-
-`slide.tidy()` runs the safe auto-fixes (`OffSlide`, `TextOverflow`,
-`OffGridDrift`); collisions, contrast and minimum size stay as warnings
-because they need a judgment call. Make it, then move on.
-
-## 10. Archetype catalog
-
-Each archetype below is complete: the code runs given the setup in §2 and
-`P`/`D` from §4. Mix 4–5 of them across a deck.
-
-```python
-def paper(slide, color):
-    slide.background.fill.solid()
-    slide.background.fill.fore_color.rgb = color
-
-def content_slide(title):
-    """Title band + body region. Every content archetype starts here."""
-    s = prs.slides.add_slide(BLANK)
-    paper(s, P.paper)
-    s.shapes.add_text(TITLE, text=title, size_pt=32, bold=True,
-                      color=P.ink, anchor="middle", margin_pt=0)
-    return s
-```
-
-**Title (dark)** — hero title, one-line subtitle, nothing else.
-
-```python
-s = prs.slides.add_slide(BLANK); paper(s, D.paper)
-s.shapes.add_text(BBox.from_inches(0.9, 2.2, W - 1.8, 1.6), text="Fotosyntéza",
-                  size_pt=54, bold=True, color=D.ink, anchor="bottom", margin_pt=0)
-s.shapes.add_text(BBox.from_inches(0.9, 3.9, W - 1.8, 0.8),
-                  text="Jak rostliny mění světlo v život · 7. ročník",
-                  size_pt=22, color=D.muted, margin_pt=0)
-```
-
-**Section divider (dark)** — number + section name, optically centred.
-
-```python
-s = prs.slides.add_slide(BLANK); paper(s, D.paper)
-s.shapes.add_text(BBox.from_inches(0.9, 2.4, W - 1.8, 0.7), text="Část 2",
-                  size_pt=20, color=D.accent, margin_pt=0)
-s.shapes.add_text(BBox.from_inches(0.9, 3.0, W - 1.8, 1.4), text="Co se děje v listu",
-                  size_pt=44, bold=True, color=D.ink, margin_pt=0)
-```
-
-**Statement** — one sentence the audience should remember; optional
-follow-up question in the accent.
-
-```python
-s = prs.slides.add_slide(BLANK); paper(s, P.paper)
-s.shapes.add_text(BBox.from_inches(1.2, 1.8, W - 2.4, 3.2),
-                  text="Každý nádech kyslíku kdysi vyrobila rostlina.",
-                  size_pt=40, bold=True, color=P.ink, anchor="middle", align="center")
-s.shapes.add_text(BBox.from_inches(1.2, 5.2, W - 2.4, 0.6),
-                  text="Proč je fotosyntéza nejdůležitější reakce na Zemi?",
-                  size_pt=18, color=P.accent, align="center")
-```
-
-**Bullets + picture** — the workhorse. List left (7/12), picture right
-(5/12), caption under the picture.
-
-```python
-s = content_slide("Co rostlina potřebuje")
-left, right = BODY.split_h([7, 5], gap=Inches(0.6))
-add_bullets(s, left, items=[
-    "Světlo — energii dodává Slunce",
-    "Vodu — kořeny ji přivádějí z půdy",
-    "Oxid uhličitý — vstupuje průduchy",
-    "Chlorofyl — zelené barvivo, které světlo zachytí",
-], size_pt=22, color=P.ink)
-add_picture_fit(s, "leaf.jpg", right, mode="contain",
-                caption="List buku · Wikimedia Commons", caption_color=P.muted)
-```
-
-**Three cards** — parallel concepts, equal weight.
-
-```python
-s = content_slide("Tři fáze v jednom listu")
-for cell, (t, b) in zip(BODY.columns(3, gap=Inches(0.5)), [
-    ("1 · Zachycení", "Chlorofyl v chloroplastech pohltí světelnou energii."),
-    ("2 · Rozklad vody", "Energie rozdělí vodu na kyslík a vodík."),
-    ("3 · Tvorba cukru", "Vodík se spojí s CO₂ a vzniká glukóza."),
-]):
-    add_card(s, cell, title=t, body=b, fill=P.surface,
-             title_color=P.ink, body_color=P.ink, title_size_pt=22, body_size_pt=18)
-```
-
-**Process** — steps with arrows, vertically centred in the body.
-
-```python
-s = content_slide("Rovnice fotosyntézy jako proces")
-horizontal_pipeline(s, BODY.inset(top=Inches(1.4), bottom=Inches(1.4)),
-                    steps=["6 CO₂ + 6 H₂O", "světlo + chlorofyl", "C₆H₁₂O₆ + 6 O₂"],
-                    accent=P.accent, fill=P.surface, text_color=P.ink,
-                    card_line=None, card_radius=10, size_pt=20)
-```
-
-**Comparison table** — two things side by side, row labels bold.
-
-```python
-s = content_slide("Fotosyntéza vs. dýchání")
-rows = [("", "Fotosyntéza", "Dýchání"), ("Kde", "chloroplasty", "mitochondrie"),
-        ("Kdy", "jen na světle", "stále"), ("Vstup", "CO₂ + H₂O", "glukóza + O₂"),
-        ("Výstup", "glukóza + O₂", "CO₂ + H₂O + energie")]
-tbl = s.shapes.add_table(len(rows), 3, *BODY.inset(bottom=Inches(1.0)), style="clean").table
-for r, row in enumerate(rows):
-    for c, val in enumerate(row):
-        tbl.cell(r, c).text = val
-tbl.format_cells(rows=0, fill=P.ink, color=P.paper, bold=True, size_pt=18)
-tbl.format_cells(rows=slice(1, None), size_pt=18, color=P.ink)
-tbl.format_cells(rows=slice(1, None), cols=0, bold=True)
-```
-
-**Two-column contrast** — "before / after", "myth / fact", "pros / cons".
-
-```python
-s = content_slide("Mýtus a skutečnost")
-a, b = BODY.columns(2, gap=Inches(0.6))
-add_card(s, a, title="Mýtus", body="Rostliny dýchají jen v noci.",
-         fill=P.surface, title_color=P.muted, body_color=P.ink, title_size_pt=20, body_size_pt=20)
-add_card(s, b, title="Skutečnost", body="Dýchají nepřetržitě; ve dne fotosyntéza převažuje.",
-         fill=P.accent_soft, title_color=P.ink, body_color=P.ink, title_size_pt=20, body_size_pt=20)
-```
-
-**Full-bleed picture with a title band** — see §7 (`mode="cover"` + a
-translucent band in `D.paper` + title in `D.ink`).
-
-**Quote** — large italic statement, attribution in `muted`.
-
-```python
-s = prs.slides.add_slide(BLANK); paper(s, P.paper)
-s.shapes.add_text(BBox.from_inches(1.4, 1.8, W - 2.8, 3.0),
-                  text="„Rostlina je stroj na slunce.“", size_pt=40, italic=True,
-                  color=P.ink, anchor="middle", align="center")
-s.shapes.add_text(BBox.from_inches(1.4, 4.9, W - 2.8, 0.6), text="— Jan Evangelista Purkyně",
-                  size_pt=18, color=P.muted, align="center")
-```
-
-**Question / activity** — one callout card, numbered prompts.
-
-```python
-s = content_slide("Zamyslete se")
-add_card(s, BODY.inset(left=Inches(1.5), right=Inches(1.5), top=Inches(0.6), bottom=Inches(0.6)),
-         title="Otázka do dvojic", body=[
-             "Proč jsou listy zelené, když černá pohltí nejvíce světla?",
-             "Co by se stalo s kyslíkem, kdyby zmizely oceánské řasy?",
-         ], numbered=True, fill=P.accent_soft, title_color=P.ink, body_color=P.ink,
-         title_size_pt=24, body_size_pt=20, pad_pt=28)
-```
-
-**Summary / closing (dark)** — three takeaways at 24 pt, then stop.
-
-```python
-s = prs.slides.add_slide(BLANK); paper(s, D.paper)
-s.shapes.add_text(BBox.from_inches(0.9, 1.4, W - 1.8, 1.0), text="Shrnutí",
-                  size_pt=36, bold=True, color=D.ink, margin_pt=0)
-add_bullets(s, BBox.from_inches(0.9, 2.6, W - 1.8, 3.6), items=[
-    "Rostliny přeměňují světlo, vodu a CO₂ na cukr a kyslík.",
-    "Děje se to v chloroplastech díky chlorofylu.",
-    "Bez fotosyntézy by nebyl kyslík ani potrava.",
-], size_pt=24, color=D.ink)
-```
-
-## 11. Look at it before you ship it
-
-Text previews cannot show a clipped title or a lopsided slide. Render the
-deck once and look:
-
-```python
-for s in prs.slides:
-    s.tidy()
-prs.save("output/deck.pptx")
-prs.render_contact_sheet("preview.png", cols=4, thumb_width=480)   # one PNG, every slide
-```
-
-What to check on the sheet, in order:
-
-1. **Fit** — every title on one line, no text touching a box edge, no
-   list shrunk to small type.
-2. **Rhythm** — titles at the same height on every content slide; the same
-   left edge everywhere; dark slides where you intended them.
-3. **Density** — no slide markedly fuller than its neighbours; no half-empty
-   slide next to a packed one.
-4. **Colour** — one accent per slide; nothing outside the palette except
-   photos.
-5. **Pictures** — undistorted, cropped where intended, captioned.
-
-Fix what you see, re-render once, save. Two passes are enough; polishing
-beyond that is time better spent on the words.
+# Slide design: making a generated deck look designed
+
+`space-aware-authoring.md` makes a deck *physically* correct — nothing
+overflows, nothing falls off the slide. This file is about the other
+half: a deck that looks like a designer laid it out. Everything here is
+built from the public API (`BBox`, `add_text`, `add_card`, `add_bullets`,
+`add_picture_fit`, `PALETTES`, `pptx2.diagrams`, `render_contact_sheet`),
+so the code samples run as written.
+
+Read this when the deck will be *seen* — a lesson, a talk, a pitch — and
+not only parsed.
+
+## The one-paragraph brief
+
+A slide that looks designed has **one idea**, **one focal point**, **one
+palette**, **one type family**, generous **air** around everything, and
+text that **fits at a readable size**. The viewer knows within a second
+where to look first. Every element on the slide is *content* (a title, a
+claim, a list, a picture, a table, a diagram, an equation) or *structure*
+(a surface, a rule, a caption); there is nothing whose only job is to
+decorate. When the content is strong the slide needs nothing else, and
+when the content is thin, ornament will not save it.
+
+## 1. Start from the content
+
+Before drawing anything, write down for each slide:
+
+- **Purpose** — one sentence: what the viewer should know or feel after it.
+- **Carrier** — the single element that delivers the purpose: a
+  statement, a short list, a picture, a table, a process, an equation.
+- **Support** — at most two more elements that help the carrier (a
+  caption, a source, a question, a picture beside a list).
+- **Archetype** — pick one from the catalog in §10. Reusing a handful of
+  archetypes is what makes a deck feel consistent.
+
+Write the archetype sequence down before you write slide 2 (a comment at
+the top of `main.py` is enough): a deck of 8–12 slides needs at least
+five distinct archetypes, consecutive slides never share one, and no
+archetype appears more than three times. Alternate dense and light
+slides: after a table or a three-card slide, give the audience a
+statement, a big number or a full-bleed picture. A deck whose content
+slides are all "title + bullets" (or all "title + three cards") is not a
+designed deck, however clean each slide is — rebuild it.
+
+Plan the pictures at the same time. A hero photo on the title (or on the
+most visual slide) plus a picture on roughly every third content slide
+is the baseline for a classroom deck; fetch them before writing the
+script so the layout is built around them rather than patched later.
+
+### Who is in the room
+
+Fix the audience before the first slide: nationality and language
+(Czech, Slovak, Polish, …), school level and age, subject. Everything
+downstream follows from it — the language and its native typography
+(diacritics, „quotation marks“, decimal comma, local dates and
+currency), the examples, names, places, institutions and curriculum
+terms the students recognise (Czech: RVP/ŠVP, Kč, Praha, Karel IV.;
+Slovak: ŠVP, Bratislava, Štúr; Polish: podstawa programowa, zł, Kraków,
+Kopernik), the vocabulary pitched to the age group. Titles are claims
+("Led je lehčí než voda"), not labels ("Hustota"); every bullet carries a
+concrete fact, number, example or question. A slide that would fit any
+school in any country is not finished.
+
+## 2. Canvas, margins, grid
+
+Work in 16:9 at 13.333 × 7.5 in, and set it explicitly:
+
+```python
+from pptx2 import Presentation, BBox
+from pptx2.util import Inches
+
+prs = Presentation()
+prs.slide_width, prs.slide_height = Inches(13.333), Inches(7.5)
+W, H = 13.333, 7.5
+BLANK = prs.slide_layouts[6]
+```
+
+Then let two regions govern every content slide:
+
+```python
+TITLE = BBox.from_inches(0.8, 0.55, W - 1.6, 0.9)     # title band
+BODY  = BBox.from_inches(0.8, 1.7,  W - 1.6, H - 2.4)  # content area (ends at 6.8")
+```
+
+- **Margins**: 0.8 in left/right, 0.55 in top, 0.7 in bottom. Nothing but
+  a full-bleed picture or a coloured background touches the slide edge.
+- **Gutters**: 0.4–0.6 in between columns and cards. Use one value per deck.
+- **Columns**: `BODY.columns(3, gap=Inches(0.5))`,
+  `BODY.split_h([7, 5], gap=Inches(0.6))`, `BODY.rows(2, gap=...)`. Every
+  card, column and picture on a slide starts at the same left edge as the
+  title, or at a grid line derived from it.
+- **Optical center**: single statements and hero titles sit slightly
+  above the geometric middle (`anchor="middle"` in a box whose top is
+  ~1.6 in and bottom ~5.4 in).
+
+## 3. Typography
+
+One family for the whole deck. Noto Sans (or whatever the environment's
+default sans is) is a fine choice; the design comes from *scale* and
+*spacing*, not from the font.
+
+| Role | Size | Weight | Notes |
+|---|---|---|---|
+| Hero title (title / section slide) | 48–60 pt | bold | 1–2 lines |
+| Statement (the one-sentence slide) | 36–44 pt | bold | ≤ 20 words, centred or left |
+| Big number | 96–120 pt | bold, `accent` | one figure + a 24 pt explanation |
+| Slide title | 32–36 pt (never below 28) | bold | one line; sentence case |
+| Body / bullets | 22–26 pt (never below 20) | regular | ≤ 6 items, ≤ 12 words each |
+| Card title | 22–24 pt | bold | |
+| Card body | 18–20 pt (never below 18) | regular | 2–4 lines |
+| Table cells | 18–20 pt | header bold | |
+| Caption / source / footer | 14 pt | regular or italic, `muted` | |
+
+Nothing on a slide is under 14 pt: a projected slide is read from the
+back of a room, and a font that only fits by going smaller is a sign of
+too much text, not of a small box. Cut words or split the slide; never
+solve density with a smaller size.
+
+Rules that make type look intentional:
+
+- **Sentence case** for titles ("Co rostlina potřebuje", not "CO ROSTLINA
+  POTŘEBUJE"). Uppercase belongs to acronyms.
+- **Left-align running text.** Centre only short statements, titles on
+  hero slides, and labels inside diagram nodes.
+- **Breathing room**: `line_spacing = 1.1–1.2` for body, `space_after =
+  8–10 pt` between bullets (`add_bullets(gap_pt=8)` does this).
+- **Zero text-box margins** when the box is already placed by the grid
+  (`margin_pt=0`) so the text edge *is* the grid line.
+- **Hierarchy by size and weight, not by colour.** Reserve colour for the
+  one thing on the slide that should pop (§4).
+- **Fit, then check**: `add_bullets` and `add_card` shrink text to the box
+  and never below 18 pt (bullets) / 16 pt (card body). If they had to
+  shrink, there is too much text — split the slide. `audit()` reports a
+  `MinFontSize` warning for any run under 12 pt.
+
+```python
+slide.shapes.add_text(TITLE, text="Co rostlina potřebuje",
+                      size_pt=32, bold=True, color=P.ink,
+                      anchor="middle", margin_pt=0)
+```
+
+## 4. Colour
+
+Pick **one** palette and use only its roles:
+
+```python
+from pptx2 import PALETTES
+
+P = PALETTES["slate"]      # slate · linen · forest · plum · ember · navy · graphite
+D = P.dark()               # same hues on a dark paper — title, section, closing slides
+```
+
+| Role | Use |
+|---|---|
+| `P.paper` | slide background (`slide.background.fill.solid(); ...fore_color.rgb = P.paper`) |
+| `P.ink` | all headings and body text |
+| `P.muted` | captions, sources, secondary labels, table grid text |
+| `P.surface` | card fills, table zebra rows, picture frames |
+| `P.line` | hairline rules and table borders |
+| `P.accent` | **one** emphasis per slide: the key number, the active step, a highlighted word, the header row |
+| `P.accent_soft` | fill of a callout card / highlighted row; text on it stays `P.ink` |
+| `P.accent_ink` | text placed *on* `P.accent` (white on a dark accent) |
+
+- **One accent per slide.** Two accents split attention; three look like a
+  toy. Cards in a row all share `P.surface`; the one to emphasise (if any)
+  gets `P.accent_soft`.
+- **Dark slides** for the title, section dividers and the closing slide
+  give the deck rhythm. Use `D = P.dark()` so the hues stay in the family.
+- **Contrast**: every text/background pair ≥ 4.5:1 (the linter reports
+  `LowContrast`); the shipped palettes satisfy this for `ink` on `paper`,
+  `surface` and `accent_soft`, for `muted` on `paper`, and for `accent_ink`
+  on `accent`.
+- **Photos bring their own colour.** Beside a photo, keep the rest of the
+  slide to `paper` / `ink` / `muted`.
+
+## 5. Surfaces: cards, callouts, panels
+
+A card is one tinted rounded rectangle with padded text inside it, and
+that is the whole card. Its silhouette stays a clean rectangle from every
+angle; emphasis comes from the tint (or, when the tint is unwanted, a
+hairline outline — one or the other, never both). Padding of 18–28 pt on
+every side gives the text room to sit rather than press against the edge.
+
+```python
+from pptx2 import add_card
+
+for cell, (title, body) in zip(BODY.columns(3, gap=Inches(0.5)), items):
+    add_card(slide, cell, title=title, body=body,
+             fill=P.surface, title_color=P.ink, body_color=P.ink,
+             title_size_pt=22, body_size_pt=18, pad_pt=22)
+```
+
+Composition rules for groups of cards:
+
+- **2–4 cards per row**, equal size, equal gap, same fill, same radius.
+  Five or more become a table.
+- **Same content shape in every card**: if one has a title and two lines,
+  all have a title and two lines. Trim copy to match.
+- **A callout** (question, key definition, "remember this") is one wide
+  card in `P.accent_soft`, inset from the body region so it reads as a
+  deliberate object, not as a background.
+- **Cards hold content, backgrounds hold cards.** A card sits directly on
+  the paper; it does not sit on another panel.
+- **Numbering** belongs *in* the title text ("1 · Zachycení"), where it
+  aligns with everything else, rather than in a separate badge shape.
+
+`add_card` returns `card.inner` (the padded box) so a picture, equation
+or diagram can be placed inside the same surface:
+
+```python
+c = add_card(slide, right_col, fill=P.surface)
+slide.shapes.add_equation(c.inner, latex=r"6\,CO_2 + 6\,H_2O \rightarrow C_6H_{12}O_6 + 6\,O_2", size_pt=28)
+```
+
+### Code listings
+
+Source code is `add_code`, never a text box with a monospace font: it
+writes one paragraph per line, verbatim, unwrapped, in a monospace face
+on a rounded editor-coloured surface, and *names the shape*
+`fika:code:<language>:<theme>:<0|1>`. An editor with a native code block
+(Fika) turns that shape into a syntax-highlighted, editable code element
+on import; PowerPoint shows a clean monospace box. A listing is a visual
+— it takes a column or the body, 18–20 pt, at most ~12 lines; longer
+code is split across slides, explained beside it, not shrunk.
+
+```python
+left, right = BODY.split_h([5, 7], gap=Inches(0.6))
+add_bullets(s, left, items=["if – když platí podmínka", "else – jinak"], size_pt=22, color=P.ink)
+add_code(s, right, code='if (odpoved == "ano")\n{\n    Console.WriteLine("Skvělé!");\n}',
+         language="csharp", theme="github-dark", size_pt=18)
+```
+
+## 6. Whitespace and alignment
+
+- Air lives in the margins and gutters — not in a half-empty body. Content
+  fills `BODY`: a row of cards spans its full width *and* height
+  (`BODY.columns(n)`), two columns split it evenly, a picture takes its
+  whole column. Never hand-size a few 3 in boxes and float them in an
+  11.7 in area. When the content is short, use fewer, larger elements
+  (two big cards, one big number, a statement) rather than small ones
+  with dead space around them.
+- **One left edge.** Title, bullets, cards and pictures share the `0.8 in`
+  margin (or a grid line derived from it). Right edges align to the
+  mirror margin.
+- **Consistent vertical rhythm.** Title band → 0.25 in gap → body. Every
+  content slide places its body at the same `BODY.top`, so flipping
+  through the deck feels steady.
+- **Balance, not symmetry.** A picture on the right with a list on the
+  left is balanced when the list's text block is roughly as tall as the
+  picture. Use `split_h([7, 5])` for text-heavy pairs, `[1, 1]` for a
+  true comparison.
+- **Empty is fine, unbalanced is not.** A slide whose right half is blank
+  because the list is short: make the list larger (24 pt), add a picture,
+  or narrow the text column and centre the pair.
+
+## 7. Pictures
+
+Photos carry emotion and specificity — one good photo does more than three
+small ones.
+
+```python
+from pptx2 import add_picture_fit
+
+# Beside text: fit inside the column, never distort, caption in muted.
+add_picture_fit(slide, "leaf.jpg", right_col, mode="contain",
+                caption="Buk lesní · Wikimedia Commons", caption_color=P.muted)
+
+# Full-bleed hero: crop to fill the whole slide, then a dark band for text.
+with slide.design_group("hero"):             # photo + band + title are one deliberate stack
+    add_picture_fit(slide, "forest.jpg", BBox.from_inches(0, 0, W, H), mode="cover")
+    band = slide.shapes.add_shape(MSO_SHAPE.RECTANGLE, *BBox.from_inches(0, H - 2.2, W, 2.2))
+    band.fill_hex(D.paper); band.line.fill.background(); band.shadow.clear()
+    band.fill.fore_color.alpha = 0.8         # translucent, photo shows through
+    slide.shapes.add_text(BBox.from_inches(0.8, H - 1.9, W - 1.6, 1.6),
+                          text="Les jako plíce planety", size_pt=44, bold=True,
+                          color=D.ink, anchor="middle", margin_pt=0)
+```
+
+- **`mode="contain"`** when the whole picture matters (diagrams, maps,
+  artworks); **`mode="cover"`** for atmosphere (backgrounds, hero slides,
+  square tiles in a grid).
+- **Picture grids**: `BODY.columns(3)` → three `mode="cover"` tiles of
+  equal size, one caption each.
+- **Caption every photo** that is not purely decorative: what it shows and
+  the source, 12–14 pt, `P.muted`.
+- Leave pictures unframed. A `P.surface` background behind a letter-boxed
+  `contain` picture is the one exception, so the empty band reads as a
+  frame rather than a gap.
+
+## 8. Tables, charts, diagrams
+
+**Tables** — structure carries the styling:
+
+```python
+tbl = slide.shapes.add_table(rows, cols, *BODY.inset(bottom=Inches(1.0)), style="clean").table
+tbl.format_cells(rows=0, fill=P.ink, color=P.paper, bold=True, size_pt=18)
+tbl.format_cells(rows=slice(1, None), size_pt=18, color=P.ink)
+tbl.format_cells(rows=slice(1, None, 2), fill=P.surface)      # zebra
+tbl.format_cells(rows=slice(1, None), cols=0, bold=True)      # row labels
+```
+
+≤ 6 body rows and ≤ 4 columns on a slide; more belongs in a handout. The
+header row is the accent of the slide — use `P.ink` or `P.accent`, not
+both.
+
+**Diagrams** — `pptx2.diagrams` recipes already space nodes and route
+arrows; feed them the palette:
+
+```python
+from pptx2.diagrams import horizontal_pipeline, cycle, hub_and_spoke, comparison_columns
+
+horizontal_pipeline(slide, BODY.inset(top=Inches(1.4), bottom=Inches(1.4)),
+                    steps=["Vstupy", "Světlo + chlorofyl", "Výstupy"],
+                    accent=P.accent, fill=P.surface, text_color=P.ink,
+                    card_line=None, card_radius=10, size_pt=20)
+```
+
+A pipeline with 3–5 steps, a cycle with 3–6 nodes, a hub with 3–6 spokes.
+Node labels are 2–4 words; explanations go in a list or a card beside the
+diagram.
+
+**Charts** — `chart.recolour([P.accent, P.muted, P.line])`, gridlines in
+`P.line`, one series in `P.accent` and the rest in greys when there is a
+story to tell. Title the chart with the takeaway ("Kyslík roste s
+osvětlením"), not the variable names.
+
+**Equations** — `slide.shapes.add_equation(bb, latex=..., size_pt=28)` or
+`paragraph.add_math(...)`; display equations at 28–36 pt, centred, with
+air around them.
+
+## 9. Overlaps: deliberate and accidental
+
+Overlap is a normal design move — text on a card, a caption on a photo, a
+translucent band over a hero image, a number disc on a timeline. The
+linter's `ShapeCollision` warning asks one question: *did you mean it?*
+
+- **Meant it** → say so, and the warning disappears:
+
+  ```python
+  with slide.design_group("hero"):           # everything created inside is one cluster
+      add_picture_fit(slide, "forest.jpg", BBox.from_slide(slide), mode="cover")
+      band = slide.shapes.add_shape(MSO_SHAPE.RECTANGLE, *BBox.from_inches(0, H - 2.2, W, 2.2))
+      slide.shapes.add_text(BBox.from_inches(0.8, H - 1.9, W - 1.6, 1.6), text="…")
+
+  slide.lint_group_overlaps(card, picture, caption)   # shapes already made, widest first
+  badge.allow_overlap_with(card)                      # exactly this pair
+  card.layer, badge.layer_above = "card", "card"      # also asserts z-order
+  ```
+
+  `add_card`, `add_picture_fit`, the components and the diagram recipes
+  already tag their own shapes, so a card and its text never trigger it.
+
+- **Did not mean it** → it is a layout bug: two text boxes crossing, a
+  card poking into its neighbour, a picture over a title. Fix the
+  geometry (use `columns()` / `split_h()` instead of hand arithmetic) and
+  re-lint.
+
+Deliberate stacking looks deliberate when the upper shape is clearly
+*inside* the lower one (inset on every side) or clearly *anchored* to one
+of its corners/edges. A shape that half-crosses a boundary reads as a
+mistake even when it is intentional.
+
+`slide.tidy()` runs the safe auto-fixes (`OffSlide`, `TextOverflow`,
+`OffGridDrift`); collisions, contrast and minimum size stay as warnings
+because they need a judgment call. Make it, then move on.
+
+## 10. Archetype catalog
+
+Each archetype below is complete: the code runs given the setup in §2 and
+`P`/`D` from §4. Mix 4–5 of them across a deck.
+
+```python
+def paper(slide, color):
+    slide.background.fill.solid()
+    slide.background.fill.fore_color.rgb = color
+
+def content_slide(title):
+    """Title band + body region. Every content archetype starts here."""
+    s = prs.slides.add_slide(BLANK)
+    paper(s, P.paper)
+    s.shapes.add_text(TITLE, text=title, size_pt=32, bold=True,
+                      color=P.ink, anchor="middle", margin_pt=0)
+    return s
+```
+
+**Title (dark)** — hero title, one-line subtitle, nothing else.
+
+```python
+s = prs.slides.add_slide(BLANK); paper(s, D.paper)
+s.shapes.add_text(BBox.from_inches(0.9, 2.2, W - 1.8, 1.6), text="Fotosyntéza",
+                  size_pt=54, bold=True, color=D.ink, anchor="bottom", margin_pt=0)
+s.shapes.add_text(BBox.from_inches(0.9, 3.9, W - 1.8, 0.8),
+                  text="Jak rostliny mění světlo v život · 7. ročník",
+                  size_pt=22, color=D.muted, margin_pt=0)
+```
+
+**Section divider (dark)** — number + section name, optically centred.
+
+```python
+s = prs.slides.add_slide(BLANK); paper(s, D.paper)
+s.shapes.add_text(BBox.from_inches(0.9, 2.4, W - 1.8, 0.7), text="Část 2",
+                  size_pt=20, color=D.accent, margin_pt=0)
+s.shapes.add_text(BBox.from_inches(0.9, 3.0, W - 1.8, 1.4), text="Co se děje v listu",
+                  size_pt=44, bold=True, color=D.ink, margin_pt=0)
+```
+
+**Statement** — one sentence the audience should remember; optional
+follow-up question in the accent.
+
+```python
+s = prs.slides.add_slide(BLANK); paper(s, P.paper)
+s.shapes.add_text(BBox.from_inches(1.2, 1.8, W - 2.4, 3.2),
+                  text="Každý nádech kyslíku kdysi vyrobila rostlina.",
+                  size_pt=40, bold=True, color=P.ink, anchor="middle", align="center")
+s.shapes.add_text(BBox.from_inches(1.2, 5.2, W - 2.4, 0.6),
+                  text="Proč je fotosyntéza nejdůležitější reakce na Zemi?",
+                  size_pt=18, color=P.accent, align="center")
+```
+
+**Bullets + picture** — the workhorse. List left (7/12), picture right
+(5/12), caption under the picture.
+
+```python
+s = content_slide("Co rostlina potřebuje")
+left, right = BODY.split_h([7, 5], gap=Inches(0.6))
+add_bullets(s, left, items=[
+    "Světlo — energii dodává Slunce",
+    "Vodu — kořeny ji přivádějí z půdy",
+    "Oxid uhličitý — vstupuje průduchy",
+    "Chlorofyl — zelené barvivo, které světlo zachytí",
+], size_pt=22, color=P.ink)
+add_picture_fit(s, "leaf.jpg", right, mode="contain",
+                caption="List buku · Wikimedia Commons", caption_color=P.muted)
+```
+
+**Three cards** — parallel concepts, equal weight.
+
+```python
+s = content_slide("Tři fáze v jednom listu")
+for cell, (t, b) in zip(BODY.columns(3, gap=Inches(0.5)), [
+    ("1 · Zachycení", "Chlorofyl v chloroplastech pohltí světelnou energii."),
+    ("2 · Rozklad vody", "Energie rozdělí vodu na kyslík a vodík."),
+    ("3 · Tvorba cukru", "Vodík se spojí s CO₂ a vzniká glukóza."),
+]):
+    add_card(s, cell, title=t, body=b, fill=P.surface,
+             title_color=P.ink, body_color=P.ink, title_size_pt=22, body_size_pt=18)
+```
+
+**Process** — steps with arrows, vertically centred in the body.
+
+```python
+s = content_slide("Rovnice fotosyntézy jako proces")
+horizontal_pipeline(s, BODY.inset(top=Inches(1.4), bottom=Inches(1.4)),
+                    steps=["6 CO₂ + 6 H₂O", "světlo + chlorofyl", "C₆H₁₂O₆ + 6 O₂"],
+                    accent=P.accent, fill=P.surface, text_color=P.ink,
+                    card_line=None, card_radius=10, size_pt=20)
+```
+
+**Comparison table** — two things side by side, row labels bold.
+
+```python
+s = content_slide("Fotosyntéza vs. dýchání")
+rows = [("", "Fotosyntéza", "Dýchání"), ("Kde", "chloroplasty", "mitochondrie"),
+        ("Kdy", "jen na světle", "stále"), ("Vstup", "CO₂ + H₂O", "glukóza + O₂"),
+        ("Výstup", "glukóza + O₂", "CO₂ + H₂O + energie")]
+tbl = s.shapes.add_table(len(rows), 3, *BODY.inset(bottom=Inches(1.0)), style="clean").table
+for r, row in enumerate(rows):
+    for c, val in enumerate(row):
+        tbl.cell(r, c).text = val
+tbl.format_cells(rows=0, fill=P.ink, color=P.paper, bold=True, size_pt=18)
+tbl.format_cells(rows=slice(1, None), size_pt=18, color=P.ink)
+tbl.format_cells(rows=slice(1, None), cols=0, bold=True)
+```
+
+**Two-column contrast** — "before / after", "myth / fact", "pros / cons".
+
+```python
+s = content_slide("Mýtus a skutečnost")
+a, b = BODY.columns(2, gap=Inches(0.6))
+add_card(s, a, title="Mýtus", body="Rostliny dýchají jen v noci.",
+         fill=P.surface, title_color=P.muted, body_color=P.ink, title_size_pt=20, body_size_pt=20)
+add_card(s, b, title="Skutečnost", body="Dýchají nepřetržitě; ve dne fotosyntéza převažuje.",
+         fill=P.accent_soft, title_color=P.ink, body_color=P.ink, title_size_pt=20, body_size_pt=20)
+```
+
+**Full-bleed picture with a title band** — see §7 (`mode="cover"` + a
+translucent band in `D.paper` + title in `D.ink`).
+
+**Quote** — large italic statement, attribution in `muted`.
+
+```python
+s = prs.slides.add_slide(BLANK); paper(s, P.paper)
+s.shapes.add_text(BBox.from_inches(1.4, 1.8, W - 2.8, 3.0),
+                  text="„Rostlina je stroj na slunce.“", size_pt=40, italic=True,
+                  color=P.ink, anchor="middle", align="center")
+s.shapes.add_text(BBox.from_inches(1.4, 4.9, W - 2.8, 0.6), text="— Jan Evangelista Purkyně",
+                  size_pt=18, color=P.muted, align="center")
+```
+
+**Question / activity** — one callout card, numbered prompts.
+
+```python
+s = content_slide("Zamyslete se")
+add_card(s, BODY.inset(left=Inches(1.5), right=Inches(1.5), top=Inches(0.6), bottom=Inches(0.6)),
+         title="Otázka do dvojic", body=[
+             "Proč jsou listy zelené, když černá pohltí nejvíce světla?",
+             "Co by se stalo s kyslíkem, kdyby zmizely oceánské řasy?",
+         ], numbered=True, fill=P.accent_soft, title_color=P.ink, body_color=P.ink,
+         title_size_pt=24, body_size_pt=20, pad_pt=28)
+```
+
+**Summary / closing (dark)** — three takeaways at 24 pt, then stop.
+
+```python
+s = prs.slides.add_slide(BLANK); paper(s, D.paper)
+s.shapes.add_text(BBox.from_inches(0.9, 1.4, W - 1.8, 1.0), text="Shrnutí",
+                  size_pt=36, bold=True, color=D.ink, margin_pt=0)
+add_bullets(s, BBox.from_inches(0.9, 2.6, W - 1.8, 3.6), items=[
+    "Rostliny přeměňují světlo, vodu a CO₂ na cukr a kyslík.",
+    "Děje se to v chloroplastech díky chlorofylu.",
+    "Bez fotosyntézy by nebyl kyslík ani potrava.",
+], size_pt=24, color=D.ink)
+```
+
+## 11. Look at it before you ship it
+
+Text previews cannot show a clipped title or a lopsided slide. Render the
+deck once and look:
+
+```python
+for s in prs.slides:
+    s.tidy()
+prs.save("output/deck.pptx")
+prs.render_contact_sheet("preview.png", cols=4, thumb_width=480)   # one PNG, every slide
+```
+
+What to check on the sheet, in order:
+
+1. **Fit** — every title on one line, no text touching a box edge, no
+   list shrunk to small type.
+2. **Rhythm** — titles at the same height on every content slide; the same
+   left edge everywhere; dark slides where you intended them.
+3. **Density** — no slide markedly fuller than its neighbours; no half-empty
+   slide next to a packed one.
+4. **Colour** — one accent per slide; nothing outside the palette except
+   photos.
+5. **Pictures** — undistorted, cropped where intended, captioned.
+
+Fix what you see, re-render once, save. Two passes are enough; polishing
+beyond that is time better spent on the words.