Posters & print
Prompting chat assistants and image generators for posters, flyers and other print pieces when you are not a designer: how designers plan a layout, the vocabulary of hierarchy, grids, typography, composition and print color, the specs a printer will ask for (sizes, resolution, bleed, PDF/X), image styles you can name, how to turn an AI image (RGB, screen resolution) into a print-ready file, and a staged prompt chain and prompt library that build every prompt on one skeleton. The general method is in Prompting fundamentals; color models and contrast are in Color theory; Gestalt, hierarchy and type scales are in UX/UI design principles; brand identity is in Branding.
How experts think
A poster has a few seconds to be noticed and one job to do. Designers fix the job, the audience, the size and the printer's specs first, then design from the outside in: trim and bleed, margins, grid, the one focal point, then everything else in order of importance.
| Question | Why it comes first | Example answer |
|---|---|---|
| What is the one message? | a poster that says five things says nothing | "Jazz night, Friday 14 March, free" |
| What should the viewer do? | sets what goes at the bottom: date, address, QR code, URL | scan the code to reserve |
| Who sees it, where, from how far? | viewing distance sets type sizes and resolution; context sets contrast | students, on a crowded corridor board, from 2–3 m |
| What is read first, second, third? | hierarchy is the order the eye takes | 1 event name, 2 date and place, 3 lineup, 4 small print |
| What size and how is it printed? | size, ratio and print method constrain everything else; changing them later means redoing the layout | A2 portrait, digital print, 50 copies |
| What are the printer's specs? | bleed, color profile, file format and minimum resolution differ by printer | 3 mm bleed, PDF/X-4, FOGRA39 profile |
| What must stay on-brand? | logos, colors and fonts may be fixed | club logo bottom right, brand orange |
| What is the image doing? | photo, illustration or pure type; it should support the message, not compete | one bold illustration of a saxophone |
Frameworks designers use:
| Framework | What it says | Ask like |
|---|---|---|
| Three levels of hierarchy | primary (seen from far away), secondary (read when interested), tertiary (details read up close) | "Sort my copy into primary, secondary and tertiary." |
| CRAP (Robin Williams, The Non-Designer's Design Book) | contrast, repetition, alignment, proximity | "Critique this layout on contrast, repetition, alignment and proximity." |
| Gestalt principles | the eye groups things by proximity, similarity, closure, continuity (details) | "Which elements group visually that shouldn't?" |
| Grid systems (the Swiss style, Josef Müller-Brockmann's Grid Systems in Graphic Design) | every element sits on a shared structure of columns, rows and margins | "Lay this out on a 6-column modular grid with 10 mm margins." |
| Squint test | blur your eyes: what still shows is your real hierarchy | "Describe what survives a squint test." |
| Outside in | trim, bleed and safe zone come before content | "Start from a 594×841 mm trim with 3 mm bleed and a 10 mm safe zone." |
Vocabulary
| Term | What it means | Say it like |
|---|---|---|
| Hierarchy | the order in which elements are seen, set by size, weight, contrast and position | "Make the date the second thing people see." |
| Focal point | the single place the eye lands first | "One focal point: the illustration, top two-thirds." |
| Visual weight | how strongly an element pulls the eye (size, darkness, saturation, detail) | "Balance the heavy image with a dark type block bottom left." |
| Negative space (white space) | empty areas, which are part of the design | "Leave a third of the poster as negative space around the headline." |
| Grid | the invisible columns, rows and margins elements align to | "Use a 12-column grid; the headline spans 10 columns." |
| Column, gutter, margin | a vertical lane; the gap between lanes; the space between content and the edge | "6 columns, 5 mm gutters, 15 mm margins." |
| Modular grid | columns plus rows, making a grid of cells | "A 4×6 modular grid; the image fills 4×4 cells." |
| Baseline grid | horizontal lines that all text sits on, so lines align across columns | "Body text on a 14 pt baseline grid." |
| Alignment | edges and centers lining up; flush left, centered, flush right, justified | "Everything flush left on one strong vertical axis." |
| Rule of thirds | splitting the frame into a 3×3 grid and placing key elements on the lines or their crossings | "Put the subject's eyes on the top-left third intersection." |
| Z-pattern, F-pattern | the paths the eye tends to follow: Z on sparse layouts (posters, ads), F on text-heavy ones | "Follow a Z-pattern: logo, headline, image, call to action." |
| Typeface vs font | the design (Helvetica) vs one file or style of it (Helvetica Bold 72 pt) | "One typeface in two weights." |
| Serif, sans serif | letters with or without small finishing strokes | "A high-contrast serif headline over a neutral sans body." |
| Display type | typefaces and sizes meant for headlines, not reading | "Display face for the title only; never for the details." |
| x-height | the height of lowercase letters like x; larger x-heights read better small | "A sans with a large x-height for the small print." |
| Kerning | adjusting the space between two specific letters | "Kern the headline by hand, especially AV, To and Ty pairs." |
| Tracking (letter-spacing) | adjusting the space evenly across a whole word or line | "Track the all-caps subhead out by +50." |
| Leading | the distance from one line's baseline to the next (named after strips of lead) | "Tight leading on the headline, about 0.9× the size." |
| Type scale | a set of sizes related by a ratio | "Use a 1.618 type scale from 24 pt body." |
| Measure | the length of a line of text | "Keep body lines under 60 characters." |
| Rag | the uneven edge of non-justified text | "Fix the rag so no line is much shorter than its neighbors." |
| Widow, orphan | a lone word or short line stranded at the end or start of a paragraph or column | "No widows in the lineup list." |
| CMYK | cyan, magenta, yellow and black (key) inks; how printers reproduce color | "Final file in CMYK, converted with the printer's profile." |
| RGB | red, green, blue light; how screens and AI tools produce color | "The AI art is RGB; convert and proof it." |
| Gamut | the range of colors a device can reproduce; CMYK's is smaller than most screens' | "Flag colors outside the CMYK gamut." |
| Spot color | a premixed ink printed as itself, not built from CMYK dots | "Print the logo orange as a spot color." |
| Pantone | the most common spot-color matching system (PMS) | "Match brand orange to a Pantone coated swatch." |
| Rich black | black built from K plus some C, M, Y for a deeper black on large areas | "Large black background in rich black; text in 100% K." |
| Total ink coverage (TAC, total area coverage) | the sum of the four ink percentages at one spot; too much smears or won't dry | "Keep total ink under the printer's limit." |
| ICC profile | a file describing a device's color behavior, used to convert accurately | "Convert with PSO Coated v3 as the printer asked." |
| Proof, soft proof | a test print; an on-screen simulation of the print | "Soft-proof for uncoated paper before ordering a hard proof." |
| PPI vs DPI | pixels per inch in an image vs dots per inch a printer lays down; often used interchangeably | "300 PPI at final size." |
| Effective resolution | an image's PPI after it is scaled in the layout | "Check effective PPI after scaling the photo to full bleed." |
| Trim | the final cut size | "Trim 18×24 in." |
| Bleed | artwork extended past the trim so cutting slightly off leaves no white edge | "0.125 in bleed on all sides." |
| Safe zone (live area) | the area inside the trim where important content must stay | "Keep text 0.25 in inside the trim." |
| Crop marks | short lines outside the page that show where to cut | "Export with crop marks and bleed." |
| Vector | artwork made of paths, sharp at any size (logos, type, flat illustration) | "Deliver the logo as vector (SVG or PDF)." |
| Raster | artwork made of pixels; blurs or pixelates when enlarged (photos, AI images) | "The AI image is raster; upscale before placing." |
| PDF/X | ISO 15930 family of PDF standards for print exchange (PDF/X-1a, PDF/X-4) | "Export PDF/X-4 with bleed and crop marks." |
| Preflight | checking a file for print problems before sending it | "Preflight for RGB images, low resolution, missing fonts and missing bleed." |
| Overprint, knockout | ink printing on top of another ink vs the ink below being removed | "Black text set to overprint." |
| Halftone | printing tones as patterns of dots | "Halftone dot texture in the shadows." |
| Duotone | an image printed in two inks | "Duotone photo in navy and orange." |
| Stock, weight | the paper; its heaviness in gsm (grams per square meter) or US pounds | "170 gsm silk coated stock." |
| Coated, uncoated | paper with or without a sealed surface; coated is sharper and more vivid | "Uncoated stock for a handmade feel; expect duller colors." |
| Finish | the surface or treatment: gloss, matte, silk, lamination, spot UV, foil | "Matte laminate, spot UV on the title." |
| Mockup | the design shown in a photo of its real setting | "Mock it up on a bus shelter to test from a distance." |
Layout and composition
Hierarchy
Three levels are enough for most posters. Every element belongs to exactly one.
| Level | Contents | Read from | Levers |
|---|---|---|---|
| Primary | the image or the headline: one thing | across a room or street | biggest size, highest contrast, most space around it |
| Secondary | what, when, where; the subhead | a few steps away | medium size, bold weight, an accent color |
| Tertiary | details, URL, sponsors, fine print | up close | small, regular weight, grouped tightly together |
Grids
| Grid | What it is | Good for |
|---|---|---|
| Manuscript | one big text area inside margins | book pages, text-only posters |
| Column | several vertical columns with gutters | flyers, posters with several text blocks |
| Modular | columns and rows forming cells | posters mixing images and text, event series |
| Baseline | horizontal lines for text to sit on | aligning text across columns |
| Hierarchical | custom zones set by content importance | expressive posters, single big image plus type |
Margins are usually larger than gutters, and the bottom margin is often a little larger than the top.
Composition
| Tool | How | Watch for |
|---|---|---|
| Rule of thirds | key elements on the third lines or their crossings | dead-center placement feels static (sometimes that is the point) |
| Single focal point | one element clearly dominates | two equally loud elements fight |
| Negative space | empty area around the focal point | filling every corner "because there is room" |
| Z-pattern | top left → top right → diagonal → bottom left → bottom right | put the call to action at the end of the Z |
| F-pattern | text-heavy layouts are scanned in horizontal bands down the left | lead each block with its key words |
| Leading lines | lines in the image point at the headline or subject | lines that lead off the edge |
| Scale contrast | one very big thing, several small ones | everything the same medium size |
Typography
| Category | Look | Use for |
|---|---|---|
| Old-style serif | soft contrast, angled stress (Garamond) | literary, classic, readable text |
| Transitional, Didone serif | high contrast, vertical stress (Baskerville, Bodoni) | fashion, elegant headlines |
| Slab serif | thick block serifs (Rockwell, Clarendon) | bold, vintage, posters |
| Grotesque, neo-grotesque sans | plain, neutral (Helvetica, Akzidenz-Grotesk) | Swiss style, information |
| Geometric sans | built from circles and lines (Futura) | modern, Bauhaus, tech |
| Humanist sans | calligraphic, open shapes (Gill Sans, Frutiger) | friendly, readable small |
| Script, blackletter, display | decorative | one word at most; never details |
| Monospace | every letter the same width | technical, retro computer looks |
| Rule | Why |
|---|---|
| At most two typefaces; get variety from weight and size | more faces look accidental |
| Kern headlines by hand | at poster sizes, default spacing gaps show (AV, To, LT) |
| Track all-caps and small text slightly looser; large display slightly tighter | caps crowd; big type looks loose |
| Headline leading about 0.85–1.1× the size; body about 1.2–1.5× | big type needs less leading than reading text |
| Use a type scale (ratio between sizes) | sizes look related, not random (type scales) |
| Strong size contrast between levels | 2–3× between primary and secondary reads as intentional |
| Check against viewing distance | a common signage rule of thumb: about 1 in (25 mm) of letter height per 10 ft (3 m) of distance for easy reading |
| Set text yourself, never in the AI image | generators misspell, and you can't edit pixels into type |
Point sizes (1 pt = 1/72 in) are typical starting points, not rules:
| Piece | Headline | Secondary | Details |
|---|---|---|---|
| Letter or A4 flyer | 36–96 pt | 18–30 pt | 9–12 pt |
| Tabloid or A3 poster | 72–150 pt | 24–48 pt | 12–16 pt |
| 18×24 in or A2 poster | 120–250 pt | 40–72 pt | 16–24 pt |
| 24×36 in or A1 poster | 150–400 pt | 48–100 pt | 20–30 pt |
Color for print
Screens mix light (RGB, additive); presses lay down ink on paper (CMYK, subtractive). Paper can't glow, so the printable range (gamut) is smaller: saturated blues, greens, oranges and neon colors from a screen usually print duller. Models and image tools work in RGB. More on color models in Color theory.
| Topic | What to do |
|---|---|
| RGB → CMYK | convert with the ICC profile the printer names (common ones: GRACoL or SWOP in the US; PSO Coated v3 (FOGRA51) or ISO Coated v2 (FOGRA39) in Europe); some printers prefer you send RGB and convert themselves, so ask |
| Soft proof | preview the conversion on screen with your editor's proof setup and turn on its gamut warning before you commit (menus in the tool syntax box under print-ready output) |
| Spot colors | for exact brand colors, metallics or fluorescents: name a Pantone; coated (C) and uncoated (U) versions of one color look different; spot colors cost extra on offset presses |
| Rich black | for large black areas, add some C, M and Y to 100% K; a common recipe is C60 M40 Y40 K100, but use your printer's; plain 100% K prints as a dark gray in big areas |
| Black text | small text and thin lines in 100% K only, or slight misalignment between plates (misregistration) makes them fuzzy |
| Total ink coverage | the limit depends on press and paper, typically around 300% on coated stock and lower on uncoated and newsprint; the printer's profile applies it |
| Paper changes color | uncoated paper absorbs ink: colors look darker and duller; tinted paper shifts everything |
| Screens lie | an uncalibrated, bright screen makes prints look dark; order a proof for anything that matters |
Print specs
Sizes
Pixel sizes are at 300 PPI at final size, without bleed. ISO A sizes all have the ratio 1:√2 (≈ 1:1.414), so a design scales cleanly from A4 to A0.
| Size | Inches | Millimeters | Pixels at 300 PPI | Ratio |
|---|---|---|---|---|
| US Letter | 8.5 × 11 | 216 × 279 | 2550 × 3300 | ≈ 1:1.29 |
| Tabloid (ledger) | 11 × 17 | 279 × 432 | 3300 × 5100 | ≈ 1:1.55 |
| 12 × 18 | 12 × 18 | 305 × 457 | 3600 × 5400 | 2:3 |
| 16 × 20 | 16 × 20 | 406 × 508 | 4800 × 6000 | 4:5 |
| 18 × 24 | 18 × 24 | 457 × 610 | 5400 × 7200 | 3:4 |
| 24 × 36 | 24 × 36 | 610 × 914 | 7200 × 10800 | 2:3 |
| 27 × 40 (US movie one-sheet) | 27 × 40 | 686 × 1016 | 8100 × 12000 | ≈ 1:1.48 |
| A6 (postcard) | 4.1 × 5.8 | 105 × 148 | 1240 × 1748 | 1:√2 |
| A5 (flyer) | 5.8 × 8.3 | 148 × 210 | 1748 × 2480 | 1:√2 |
| A4 | 8.3 × 11.7 | 210 × 297 | 2480 × 3508 | 1:√2 |
| A3 | 11.7 × 16.5 | 297 × 420 | 3508 × 4961 | 1:√2 |
| A2 | 16.5 × 23.4 | 420 × 594 | 4961 × 7016 | 1:√2 |
| A1 | 23.4 × 33.1 | 594 × 841 | 7016 × 9933 | 1:√2 |
| A0 | 33.1 × 46.8 | 841 × 1189 | 9933 × 14043 | 1:√2 |
With bleed, add it to every side: 24 × 36 in becomes 24.25 × 36.25 in (7275 × 10875 px); A3 becomes 303 × 426 mm (3579 × 5031 px); A2 becomes 426 × 600 mm (5031 × 7087 px).
File setup
| Spec | Typical value | Notes |
|---|---|---|
| Resolution | 300 PPI at final size | large posters seen from a distance often print fine at 150–200 PPI; ask the printer |
| Bleed | 0.125 in (US) or 3 mm (Europe) on every side | some large-format printers ask for more, such as 5 mm |
| Safe zone | text and logos at least 0.125–0.25 in or 3–5 mm inside the trim; more on big posters | cutting can drift slightly |
| Color | CMYK with the printer's profile, or RGB if the printer converts | ask; spot colors named exactly as in the swatch book |
| Type | live text in a layout app; fonts embedded in the PDF | outlining fonts is a fallback some printers ask for |
| Logos, flat art | vector | raster logos blur at poster size |
| Lines | at least 0.25 pt | thinner hairlines can disappear |
| File | PDF/X-1a (everything flattened to CMYK and spot) or PDF/X-4 (allows live transparency and color-managed RGB) | crop marks offset so they sit outside the bleed |
| Preflight | run it before sending | InDesign, Illustrator, Affinity and Acrobat have preflight tools |
Office and home printers usually can't print to the edge of the sheet (an unprintable margin of about 0.25 in, or 6 mm, is common), so for a piece printed on one, skip the bleed and keep everything inside that margin.
Resolution and viewing distance
| Piece | Typical viewing distance | Resolution at final size |
|---|---|---|
| Cards, flyers, postcards, books | in hand, 30–50 cm | 300 PPI |
| Small posters (up to A3 or 11 × 17 in) | arm's length | 300 PPI |
| Large posters (A1, A0, 24 × 36 in) | 1–3 m | 150–300 PPI |
| Banners and billboards | several meters and more | much lower; follow the printer's spec exactly |
A rough check: with normal (20/20) vision you resolve about 1 arcminute, which works out to needing about 87 ÷ distance in meters PPI (about 3,438 ÷ distance in inches). At 1 m that is ≈ 87 PPI, at 30 cm ≈ 290 PPI. Extra resolution beyond that is invisible, but printers' minimums still apply.
Paper and finishes
| Stock or finish | Look and feel | Use for |
|---|---|---|
| Gloss coated | shiny, vivid, sharp; glare under lights | photo posters, bright colors |
| Silk or satin coated | low sheen; vivid without much glare | most posters and flyers |
| Matte coated | no shine, softer contrast | text-heavy pieces, elegant looks |
| Uncoated | natural, absorbent, writable; colors duller and darker | handmade looks, stationery, risograph |
| Weight | flyers ≈ 130–170 gsm (≈ 90–115 lb text); posters ≈ 150–200 gsm; cards ≈ 300–400 gsm (≈ 110–150 lb cover) | heavier feels premium and costs more |
| Lamination | a plastic film: gloss, matte or soft-touch | durability, handled pieces |
| Spot UV | a glossy varnish on selected areas | making a title or image shine |
| Foil, embossing, die-cut | metallic stamp; raised or pressed shapes; custom cut shapes | premium pieces; set up as separate vector layers |
Print methods
| Method | Good for | Constraint |
|---|---|---|
| Digital (toner or inkjet) | short runs, fast, variable data | CMYK only on most machines |
| Offset lithography | long runs, low cost per copy, spot colors | setup cost; minimum quantities |
| Large-format inkjet | single posters, banners, art prints | check the printer's profile and maximum size |
| Screen printing | gig posters, bold flat colors, special inks | one screen per color; design in separated flat colors |
| Risograph | zine and art-print looks, grainy spot colors, slight misregistration | limited ink colors; design each color as a grayscale layer |
Photography and illustration styles
Name styles by their visible qualities; a model draws on that vocabulary. Avoid naming living artists: describe what you like about their work instead (and check the tool's rules).
| Style | Visual qualities | Say it like |
|---|---|---|
| Swiss (International Typographic Style) | grids, sans serif, asymmetric layouts, photos, lots of white | "Swiss-style poster: grid, flush-left Helvetica-like sans, one photo, red accent." |
| Bauhaus, Constructivism | geometric shapes, diagonals, primary colors, bold type | "Constructivist composition: diagonals, red and black, geometric blocks." |
| Art Nouveau | flowing organic lines, decorative borders, ornate lettering | "Art Nouveau frame of vines and flowers, muted gold and green." |
| Art Deco | symmetry, geometric ornament, metallic accents, stepped forms | "Art Deco travel poster: symmetrical, sunburst, gold and navy." |
| 1930s travel and park posters | flat screen-printed color, simplified landscapes, strong silhouettes | "Vintage screen-printed national park poster, 5 flat colors." |
| 1960s psychedelic | warped lettering, vibrating complementary colors | "1960s psychedelic concert poster colors and swirling shapes, no text." |
| Mid-century modern | simplified shapes, limited palette, playful | "Mid-century illustration, flat shapes, mustard, teal and orange." |
| Minimalist | one element, huge negative space | "Minimalist poster: one small object, 80% empty space." |
| Screen print, risograph | flat spot colors, grain, slight misregistration | "Two-color risograph look, fluorescent pink and blue, visible grain." |
| Linocut, woodcut | carved lines, high contrast, one or two inks | "Linocut illustration, black on cream, bold carved lines." |
| Flat vector | clean shapes, no texture, solid fills | "Flat vector illustration, 6 colors, no gradients." |
| Collage | cut paper, mixed photos, torn edges | "Paper collage of cut photos and torn colored paper." |
| Halftone, pop art | visible dots, bold outlines, primaries | "Halftone dot shading, bold outlines, comic colors." |
| Studio product photo | clean background, controlled light | "Product photo on seamless gray, softbox from left, soft shadow." |
| Editorial portrait | natural or dramatic light, shallow depth of field | "Portrait, 85 mm lens, f/2, window light, muted colors." |
| Documentary, street | candid, available light, grain | "Candid street photo, 35 mm, natural light, film grain." |
| 3D render | clay or plastic materials, soft global lighting | "Soft 3D clay render, pastel colors, studio lighting." |
Photo vocabulary that steers generators: focal length (24 mm wide, 50 mm normal, 85–135 mm portrait, macro); depth of field (shallow with a blurred background, or deep); lighting (softbox, hard flash, rim light, golden hour, high-key bright or low-key dark); angle (eye level, low angle, top-down flat lay).
Print-ready output from AI image tools
Image generators output RGB at screen resolution (often about 1–4 megapixels, varying by tool and setting) in a fixed set of aspect ratios, without bleed, and with no guarantee the text, hands or small details are right. A 24 × 36 in poster at 300 PPI needs about 78 megapixels. So treat the AI image as raw art and finish the file yourself.
Fix the size and ratio first
Pick the final size, then generate at the closest ratio the tool offers: 2:3 for 24 × 36 or 12 × 18 in, 3:4 for 18 × 24 in, 4:5 for 16 × 20 in. For ISO A sizes (1:1.414) generate slightly wider or taller and crop, or extend the canvas.Generate art without text
Ask for "no text, no letters, no logos" and for empty space where the type will go ("empty sky in the top third"). Set every word yourself in a layout app.Extend for bleed
Use outpainting or generative expand to add canvas beyond the trim, so the art runs into the bleed. Nothing important within the bleed or safe zone.Upscale
Use an upscaler (built into some generators and editors, or dedicated ones such as Topaz Gigapixel, Magnific or the open-source Upscayl) to reach the target pixels. Upscale in steps of 2× and inspect each step; upscalers invent detail.Inspect at 100%
Check hands, faces, eyes, teeth, edges, repeating patterns, melted objects, stray letters, and seams from outpainting. Retouch or regenerate those areas.Convert and proof color
Convert to CMYK with the printer's profile (or send RGB if they convert), soft-proof, and fix colors that fall outside the gamut.Lay out and set type
Place the image in a layout app (InDesign, Illustrator, Affinity, Scribus, or Canva with its print export), add the type on a grid, keep it inside the safe zone.Vectorize flat art if needed
Logos and flat illustrations can be traced to vectors in a vector editor (Illustrator, Inkscape, Affinity) or a tracing service such as Vectorizer.AI; clean up the paths by hand, and check the result still reads about 15 mm wide and in one color.Export, preflight, proof
Export PDF/X with bleed and crop marks, run preflight, and order a proof print for anything expensive.
How much to upscale (starting from a 2:3 image):
| Target | Pixels needed | From 1024 × 1536 | From 2048 × 3072 |
|---|---|---|---|
| 12 × 18 in at 300 PPI | 3600 × 5400 | ≈ 3.5× | ≈ 1.8× |
| 24 × 36 in at 150 PPI | 3600 × 5400 | ≈ 3.5× | ≈ 1.8× |
| 24 × 36 in at 300 PPI | 7200 × 10800 | ≈ 7× | ≈ 3.5× |
| A3 at 300 PPI (after cropping) | 3508 × 4961 | ≈ 3.4× | ≈ 1.7× |
Text. Some generators render short text well (recent versions of Ideogram, Recraft, and the image tools in ChatGPT and Gemini), which is useful for quick mockups and layout ideas. For the final file, set type yourself: you can proofread it, kern it, change it, and it prints sharp as vector. Recraft can also output vector (SVG) images. Capabilities change quickly; test before relying on them.
Art prints. Photo labs that print art prints often want a high-quality JPEG or TIFF in sRGB or Adobe RGB rather than a CMYK PDF/X; press printers want PDF/X. Ask which, and soft-proof with the lab's profile.
Brief the model
The Slot column says where each input goes in the prompt skeleton that every full prompt below uses.
| Input | Slot | Why it matters | Example |
|---|---|---|---|
| Purpose and one message | <context>, <task> | sets the hierarchy | "get students to a free jazz night" |
| Audience and setting | <context> | sets tone, contrast and size of type | "students; corridor notice board, cluttered" |
| Viewing distance | <context> | sets type sizes and resolution | "read from 2–3 m" |
| Size, orientation, bleed | <context>, <requirements> | everything is laid out to it | "A2 portrait, 3 mm bleed" |
| Print method and paper | <context> | changes color, detail and file setup | "digital print on 170 gsm silk" |
| All the copy, in priority order | <materials> | the model can't guess your details | headline, date, place, lineup, URL, sponsors |
| The printer's file requirements | <materials> | bleed, profile and PDF/X version differ by printer | the "file setup" page from their website, pasted |
| Brand assets | <context>, <requirements> | fonts, colors, logo rules | "brand orange #E8590C, Inter, logo bottom right" |
| Style references | <examples> | turns taste into words; one you dislike, with why, sharpens it | "1960s Blue Note album covers: big type, one photo, two colors" |
| Image role | <context> | photo, illustration, type-only | "one illustration; it carries the mood" |
| Constraints and standards | <requirements>, <quality_bar> | what to avoid, and the standard to check against | "no faces; must work in grayscale; PDF/X-4; three levels of hierarchy" |
| Output format | <output_format> | gets you a spec, not an essay | "layout spec: grid, sizes in pt, positions in mm" |
What to ask for
| Stage | Deliverable | Contents |
|---|---|---|
| Concept | creative brief | goal, audience, message, tone, references, mandatories |
| 3 concepts | each: idea in one line, image description, type approach, palette | |
| copy hierarchy | the text sorted into primary, secondary, tertiary, edited to fit | |
| Image | image prompts | one per concept, at the right ratio, with no text and space for type |
| art direction notes | lighting, palette, composition rules for regenerations | |
| Layout | layout spec | trim, bleed, margins, grid, each element's position, size and font |
| type spec | typefaces, weights, sizes, leading, tracking per level | |
| color spec | palette in CMYK (and Pantone if spot), rich black recipe | |
| Prepress | print spec sheet | size, bleed, resolution, profile, PDF standard, stock, finish, quantity |
| preflight checklist | what to check before sending | |
| printer email | a short message confirming specs and asking for their profile |
The prompt chain
Four prompts, run in order, for the field's main deliverable: a print-ready poster brief and layout spec, with the print spec sheet a printer needs before accepting the file. Stage 1 gets the brief out of you, stage 2 makes you choose a concept, stage 3 writes the spec, and stage 4 checks it in a fresh chat, because a model reviewing its own draft shares its blind spots.
1. Interview
Run it first, when all you have is a reason for a poster: it asks what a designer and a printer would ask.
<context>
I need a poster and I'm not a designer. What it's for: [one or two
sentences, e.g. "a free student jazz night; I want people to
reserve"]. What I have: [e.g. "the copy in a doc, a club logo, a
print shop nearby"].
</context>
<task>
Interview me before anything is designed. I'll use my answers to
brief a poster layout and a print spec a printer can work from.
</task>
<requirements>
- At most 10 questions, grouped under: Message, Viewer, Copy, Format
and printing, Brand and image.
- Cover the one message; the action; where it hangs and from how far
it is read; every piece of copy; size and orientation; print
method, paper, quantity and deadline; the printer's specs (bleed,
safe zone, color profile, PDF/X version, minimum resolution);
brand rules; the image's job.
- For each question: why it matters, and 2–3 typical answers.
</requirements>
<output_format>
Numbered questions under the five group headings, each with a
"Why:" line and a "Typical:" line.
</output_format>
<process>
Ask the questions and stop. Don't design anything yet. When I have
answered, write my answers back as a brief of at most 12 bullets and
list what I still need to ask the printer.
</process>2. Options
Run it with the confirmed brief: the same copy can make very different posters, and the choice is yours.
<materials>
[paste the brief from stage 1]
</materials>
<task>
Propose 3 materially different layout concepts for this poster, so I
can pick one to specify.
</task>
<requirements>
- The concepts differ in what carries the message, for example:
type-led on a Swiss-style modular grid; image-led with one dominant
photo or illustration; a limited-color screen-print or risograph
look.
- For each: the idea in one line; the focal point; the hierarchy
(what is read 1st, 2nd, 3rd); the grid; the type approach; the
palette (CMYK-safe, plus any spot colors); the image's job and
whether it needs AI art, a photo or none; the risk at the viewing
distance and in a cluttered setting; what drives the print cost.
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Each concept has one focal point and survives a squint test.
- The one message reads from the stated distance in every concept.
- The three would look different at thumbnail size.
</quality_bar>
<output_format>
A comparison table with one column per concept, then a
recommendation in 3 sentences: which, why, and what would change
your mind.
</output_format>
<process>
Stop after the recommendation and wait for my choice. End with a
list of the assumptions you made.
</process>3. Produce
Run it with the concept you picked: it writes the brief, the layout spec and the print spec in one pass, so the design and the file agree.
<materials>
[paste the brief from stage 1 and the concept you chose]
[paste all the copy, in priority order]
[paste the printer's file requirements, if you have them]
</materials>
<context>
Poster: [purpose and action, e.g. "free jazz night; scan to
reserve"].
Hangs: [where and the viewing distance, e.g. a crowded corridor
board, read from 2–3 m].
Size: [size: A3 | A2 | A1 | 18 × 24 in | 24 × 36 in], [orientation: portrait | landscape].
Printing: [print method: digital | offset | large-format inkjet | screen print | risograph] on [paper, e.g. 170 gsm silk], [quantity, e.g. 50 copies].
Brand: [colors, fonts and logo rules, e.g. "brand orange #E8590C,
Inter, logo bottom right", or "none"].
Layout app: [app: InDesign | Affinity | Illustrator | Scribus | Canva]
</context>
<examples>
Posters I like: [1–3, each with what to take, e.g. "1960s jazz
record covers: huge type, one photo, two colors plus black"]
One I don't: [one, with why, e.g. "five fonts and no focal point"]
</examples>
<task>
Write the poster brief, a layout spec and a print spec sheet, so I
can build the file in my layout app and send it to the printer.
</task>
<requirements>
- Hierarchy in three levels (primary, secondary, tertiary); every
piece of copy in exactly one; one focal point.
- Outside in: trim, bleed, safe zone, then a column or modular grid
with margins, gutters and a slightly larger bottom margin, in mm
(or inches). Place each element by column and row.
- Type: at most two typefaces (with free alternatives), sizes from a
type scale with its ratio stated, leading and tracking per level.
Check each level against the viewing distance: about 25 mm of
letter height per 3 m.
- Color: the palette in CMYK and hex, Pantone if spot; rich black for
large areas, 100% K for small text and hairlines; total ink under
the printer's limit.
- Print spec: trim; bleed (3 mm or 0.125 in unless the printer says
otherwise); safe zone; effective resolution (300 PPI at final
size, or 150–200 if the printer accepts it for a large poster);
ICC profile; PDF/X-4 or PDF/X-1a; crop marks; fonts embedded.
- If there is an AI image: the ratio to generate at, the upscale
factor to the pixels needed, and empty space where the type goes.
- Never invent numbers, names, quotes or sources: write [CHECK WITH
PRINTER] for any spec the printer must confirm, and [NEEDS SOURCE]
where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- The squint test: the primary still leads, then the secondary.
- Every type size is computed from the viewing distance, not
guessed.
- Text, logo and QR code sit inside the safe zone; art runs into the
bleed.
- The spec sheet answers every question a printer asks before
accepting a file.
</quality_bar>
<output_format>
1. The brief, at most 10 bullets. 2. A copy hierarchy table (level |
copy). 3. A layout table (element | columns and rows | position from
the trim in mm | size | typeface, weight, size in pt, leading,
tracking | color). 4. A color table (name | CMYK | hex | Pantone).
5. The print spec sheet as a table. 6. The image prompt, if any.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>4. Review
Run it in a fresh chat with the brief, the spec and a proof or PDF export: the critic should not be the author.
<materials>
<brief>
[paste the brief]
</brief>
<draft>
[paste the layout and print spec, and attach a proof or the PDF
export, or describe it]
</draft>
</materials>
<task>
Review this poster and its print spec before I send it to the
printer. This is a fresh chat: judge only what is here.
</task>
<requirements>
- Score each check 1–5 with one line of evidence: hierarchy and the
squint test; the one message at the viewing distance; grid and
alignment; type; negative space; color (contrast, CMYK risk,
grayscale); image artifacts at 100%; specs (trim, bleed, safe zone,
effective PPI, profile, PDF/X, black text); rights.
- List every issue that would make a printer reject the file or
print it badly.
- Don't redesign it or rewrite the copy. Don't praise it.
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Every problem points to an element and a measurement.
- Every fix is one concrete instruction ("set the date at 96 pt",
"move the QR code 6 mm in from the trim").
</quality_bar>
<output_format>
A score table (check | score | evidence), then at most 7 fixes
ranked by effect, with issues a printer would reject first.
</output_format>
<process>
End with a list of the assumptions you made.
</process>Goals and loops
In an agent that runs commands and checks its own work, such as Claude Code, the Produce stage can run until it
measurably passes: /goal keeps starting turns until a condition holds, and /loop repeats a prompt or skill on
an interval. The checker reads only the conversation, so every condition below names a check whose output Claude
prints, the constraints that must survive, and a bound. How to write your own:
Goals and loops.
| Task | Completion condition |
|---|---|
| Print file check | /goal magick identify -format "%w x %h %[colorspace] %x" poster.tif prints 5031 x 7087 CMYK at 300 PPI (A2 with 3 mm bleed), and pdffonts poster.pdf lists every font as embedded; or stop after 8 turns |
| Copy hierarchy | /goal the copy deck has exactly three levels, a headline of at most seven words, and every date, time and place matches event.md (print the comparison); or stop after 6 turns |
| Art prompts | /goal four generator prompts, each with the 5:7 aspect ratio for A sizes, no text in the image, and the palette from the brief quoted; or stop after 4 turns |
Loops that fit: none, usually: print work ends at the file. A /loop only helps while waiting on something external, such as a print shop's upload portal or proof status.
What a checker can't judge: whether it reads from three meters and looks right on paper. Print a proof at size and stand back.
Prompt library
One entry per thing a designer or prepress technician hands over. Quick is the minimum that still works; Full
fills in the skeleton. Replace every […] and delete the optional slots you leave empty.
Copy hierarchy
Stage: produce · Use when: you have all the text and need it cut and sorted before any layout.
Quick
Sort this poster copy into primary, secondary and tertiary: [paste].
The one message: [e.g. "free jazz night, Friday"]. Cut what doesn't
serve it and tell me what you cut; keep every fact exactly as given.
Check that the primary is one thing a passer-by gets in 3 seconds.Full
<materials>
[paste all the copy, as you have it]
</materials>
<context>
Piece: [piece: poster | flyer | postcard | banner], [size, e.g. A2 portrait].
Goal: [the action, e.g. "scan the QR code to reserve"].
Seen: [where and from how far, e.g. a corridor board, 2–3 m].
Mandatories: [what must appear, e.g. sponsor logo, accessibility
note, age limit].
</context>
<task>
Edit the copy into a three-level hierarchy, so the layout can give
each level its size.
</task>
<requirements>
- Three levels: primary (seen from across the room: one thing),
secondary (what, when, where), tertiary (details read up close).
- Cut or shorten anything that doesn't serve the goal; list each cut
with a reason.
- Keep facts exactly: dates, times, names, prices, URLs.
- Rewrite only for length and clarity; keep the organizer's names
and terms.
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where a fact is missing (a time, an address).
</requirements>
<quality_bar>
Before answering, check your draft against:
- The primary is one message of at most about 6 words.
- The secondary answers what, when and where.
- No fact changed between my copy and yours.
</quality_bar>
<output_format>
A table (level | copy | word count), then the list of cuts.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>Poster art prompts
Stage: produce · Use when: you need the main image from a generator, at a print ratio, with no text and room for the type.
Quick
Write 4 image-generator prompts for poster art: [subject, e.g. a
saxophone player's hands], [style, e.g. vintage screen-print look, 5
flat colors], at [ratio: 2:3 | 3:4 | 4:5] portrait, no text or logos,
with empty space in [where, e.g. the top third] for the headline.
Check that the art never competes with the headline for focus.Full
<materials>
[paste the chosen concept and the layout spec, if you have them]
</materials>
<context>
Poster: [purpose, e.g. a free student jazz night], [size, e.g. A2
portrait].
Printing: [print method: digital | offset | large-format inkjet | screen print | risograph]
Image tool: [tool, e.g. Midjourney, Firefly, Ideogram or a chat
assistant's image mode].
Where the type goes: [e.g. headline in the top 40%, details along
the bottom 15%].
</context>
<examples>
Images I like: [1–3, each with what to take: palette, lighting,
texture or composition]
One I don't: [one, with why, e.g. "glossy stock-photo look"]
</examples>
<task>
Write prompts for the poster's main image, so I can generate it,
pick one and finish it for print.
</task>
<requirements>
- Generate at the ratio nearest the trim that the tool offers (2:3
for 24 × 36 in, 3:4 for 18 × 24 in, 4:5 for 16 × 20 in; for A
sizes, 2:3 then crop or 3:4 then extend).
- Each prompt: subject and action; style by visible qualities, never
a living artist's name; palette; lighting; composition with the
empty area named; mood; "no text, no letters, no logos"; art to all
four edges; aspect ratio.
- Vary one thing per prompt: lighting, viewpoint, palette or style.
- A palette that survives CMYK: no neon screen colors unless a
fluorescent spot ink is planned.
- The upscale factor from the tool's output to the pixels needed at
final size plus bleed.
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- The image supports the message instead of fighting the headline
for the focal point.
- The empty area is big enough for the primary level.
- No prompt asks for text, hands at work or exact counts.
</quality_bar>
<output_format>
The 4 prompts in separate text blocks, each with a one-line note on
what it varies; then the ratio, the crop or extend step and the
upscale factor in 3 lines.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>Type spec
Stage: produce · Use when: you don't know which typefaces or sizes to use, or the layout feels flat.
Quick
Suggest 3 type pairings (a display face and a text face, each free on
Google Fonts, plus a paid alternative) for a [mood, e.g. warm, retro]
poster. For the one I pick, give sizes in pt per hierarchy level on a
[size, e.g. 24 × 36 in] poster read from [distance, e.g. 3 m], with
leading and tracking. Check every size against about 25 mm of letter
height per 3 m.Full
<materials>
[paste the copy hierarchy]
</materials>
<context>
Poster: [size, e.g. 24 × 36 in], read from [distance, e.g. 1–3 m].
Mood: [e.g. "late-night, warm, a little retro"].
Brand fonts: [fonts, e.g. Inter, or "none"].
Category I lean toward: [category: old-style serif | Didone | slab serif | grotesque sans | geometric sans | humanist sans]
</context>
<examples>
Type I like: [1–3 posters, covers or sites, each with what to take,
e.g. "huge condensed sans, tight leading"]
Type I don't: [one, with why, e.g. "script caps on the details"]
</examples>
<task>
Write a type spec for the poster, so every hierarchy level has a
face, a size, leading and tracking.
</task>
<requirements>
- At most two typefaces; variety from weight and size.
- Sizes from a type scale: state the ratio (e.g. 1.5 or 1.618) and
the base size.
- 2–3× size contrast between primary and secondary.
- Headline leading about 0.85–1.1× the size, body about 1.2–1.5×;
all-caps and small text tracked slightly looser, big display type
slightly tighter.
- The letter pairs to kern by hand in the headline.
- Each level checked against the viewing distance: about 25 mm of
letter height per 3 m, using cap height (roughly 0.7× the point
size, which varies by face).
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where support is missing, and [CHECK LICENSE] for any font
license I must confirm.
</requirements>
<quality_bar>
Before answering, check your draft against:
- No level needs a third typeface.
- Every size passes the distance check or is marked "read up close".
- The headline still reads at thumbnail size.
</quality_bar>
<output_format>
The 3 pairings as a table (display | text | free source | paid
alternative | why), then the spec as a table (level | typeface and
weight | size in pt | cap height in mm | leading | tracking | reads
from), then the kerning pairs.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>Color spec
Stage: produce · Use when: you need a palette that prints the way it looks, with the CMYK, Pantone and black recipes the printer will ask about.
Quick
Turn these screen colors into a print palette: [paste hex values].
Printing [print method: digital | offset | screen print] on [paper, e.g. uncoated 150 gsm].
Give CMYK builds, which colors will shift or fall out of gamut,
Pantone matches to check in a swatch book, and a rich black recipe.
Check that text and background still contrast in grayscale.Full
<materials>
[paste the colors you have: hex values, a brand guide or the image]
</materials>
<context>
Printing: [print method: digital | offset | large-format inkjet | screen print | risograph] on [paper: coated gloss | coated silk | coated matte | uncoated].
Printer's profile: [e.g. PSO Coated v3 (FOGRA51), GRACoL, or "not
asked yet"].
Brand rules: [e.g. "the orange must match Pantone Orange 021 C", or
"none"].
</context>
<task>
Write a color spec for the poster, so the printed colors match what
I approved on screen as closely as the press allows.
</task>
<requirements>
- Each color: its role (background, primary type, accent), hex, CMYK
build, and whether it falls outside the CMYK gamut and how it will
shift.
- Spot colors only where they earn their cost (an exact brand color,
a metallic, a fluorescent); coated (C) and uncoated (U) swatches
differ.
- Rich black for large areas (a common recipe is C60 M40 Y40 K100;
confirm the printer's); 100% K for small text and thin lines.
- Total ink under the printer's limit (typically around 300% on
coated stock, lower on uncoated).
- How uncoated or tinted paper will change the palette.
- Never invent numbers, names, quotes or sources: write [CHECK WITH
PRINTER] for the profile, ink limit and black recipe, and [NEEDS
SOURCE] where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Text and background still contrast when converted to grayscale.
- No small text uses rich black or a four-color build.
- Every out-of-gamut color has a stated fallback.
</quality_bar>
<output_format>
A table (role | hex | CMYK | Pantone to check | gamut risk |
fallback), the black recipes, and 3 questions to ask the printer.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>Screen-print separations
Stage: produce · Use when: the look should be flat, bold and limited-color, or the poster will really be screen printed or risographed.
Quick
Plan a [inks: 2 | 3 | 4]-ink screen-print gig poster of [subject, e.g. a
surf-rock band]: the image prompt (flat colors, no gradients, no text,
an empty band for the name), how to vectorize it and separate it into
one layer per ink, and what to ask the printer about trapping. Check
that every element is one flat ink or an intended overlap.Full
<materials>
[optional: paste or attach the sketch or AI image you have]
</materials>
<context>
Poster: [event or subject, e.g. a gig poster for a surf-rock band],
[size, e.g. 18 × 24 in].
Method: [method: screen print | risograph], [inks: 2 | 3 | 4] inks.
Ink colors: [e.g. fluorescent pink, teal and black, or "help me
choose"].
Tools: [e.g. an image generator, then Illustrator or Inkscape].
</context>
<examples>
Gig posters I like: [1–3, each with what to take, e.g. "two inks
overlapping to make a third"]
One I don't: [one, with why, e.g. "photographic detail that won't
separate"]
</examples>
<task>
Plan the artwork and color separations for a limited-ink print, so
the printer can make one screen (or master) per ink.
</task>
<requirements>
- The image prompt: flat shapes; exactly the stated inks; no
gradients or photographic detail; slight paper texture at most;
an empty band for the type; no text.
- Vectorize with a color limit equal to the ink count; clean the
paths; one layer per ink, named after the ink.
- For a risograph: each ink as a grayscale layer; plan for slight
misregistration and for overlaps that make extra colors.
- Type set by me as vector and assigned to one ink.
- Never invent numbers, names, quotes or sources: write [CHECK WITH
PRINTER] for trapping, minimum line weight and ink choices, and
[NEEDS SOURCE] where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Every element is one ink or an intended overlap.
- Thin lines and small type are thick enough for the method.
- The design still reads if registration drifts by about a
millimeter.
</quality_bar>
<output_format>
The image prompt in a text block, a separation table (ink | what it
prints | layer name | notes), and the steps from image to files.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>Preflight checklist
Stage: produce · Use when: the layout is done and you are about to export and send the file.
Quick
I'm printing [quantity and size, e.g. 50 × A2] by [print method: digital | offset | large-format inkjet].
Files: [e.g. an AI image upscaled to 5031 × 7087 px, RGB; type live in
InDesign]. Write a spec sheet (trim, bleed, safe zone, effective PPI,
profile, PDF/X, crop marks) and a preflight checklist, and flag what
in my setup will cause a problem.Full
<materials>
[paste the printer's file requirements, or write "none yet"]
[paste the layout spec]
</materials>
<context>
Job: [quantity and size, e.g. 50 × A2 portrait] on [stock and
finish, e.g. 170 gsm silk, no lamination].
Method: [print method: digital | offset | large-format inkjet | screen print | risograph]
Files: [what's in the layout, e.g. "an AI image upscaled to 5031 ×
7087 px, RGB; type live in InDesign; logo as SVG; QR code as PNG"].
Export from: [app and version, e.g. InDesign 2025].
</context>
<task>
Write a print spec sheet and a preflight checklist for this job, so
the file passes the printer's checks the first time.
</task>
<requirements>
- Spec sheet: trim; bleed; safe zone; effective resolution of each
image at its placed size; color mode and profile; PDF/X version;
crop mark offset; font embedding; total ink; rich black and black
text.
- Checklist items each testable in the app's preflight or by eye:
RGB or low-resolution images, missing fonts, missing bleed, text in
the bleed, hairlines under 0.25 pt, rich black on small text, spot
colors named as in the swatch book.
- The QR code: dark on light, a quiet zone of 4 modules around it,
tested from a print at real size.
- Never invent numbers, names, quotes or sources: write [CHECK WITH
PRINTER] for any value only the printer can confirm.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Every item says how to check it and what passes.
- The effective PPI is computed for each image at its placed size.
- Nothing contradicts the printer's requirements.
</quality_bar>
<output_format>
The spec sheet as a table (spec | value | from the printer or a
default), then the checklist as checkboxes grouped by stage:
document setup, images, type, color, export.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change
your answer; otherwise go on. End with a list of the assumptions you
made.
</process>Printer email
Stage: produce · Use when: you are starting a job with an unfamiliar printer and want their specs before you design.
Quick
Write a short, polite email to a print shop. Job: [quantity and size,
e.g. 50 × A2] on [stock, e.g. 170 gsm silk], needed by [date, e.g. 3 March]. Ask for
bleed and safe zone, minimum resolution, color profile or RGB/CMYK,
PDF/X version, rich black and ink limit, proofing and file delivery.
Under 150 words; check that each question can be answered with a value.Full
<materials>
[optional: paste the file guidelines from the printer's website]
</materials>
<context>
Job: [quantity, e.g. 50] × [size, e.g. A2 portrait] on [stock, e.g. 170 gsm silk], [finish: none | matte laminate | gloss laminate | spot UV | foil].
Method: [print method: digital | offset | large-format inkjet | your recommendation]
Deadline: [date and whether it can move, e.g. 3 March, fixed].
Me: [who, e.g. a student society, first job with this shop].
</context>
<task>
Write an email to the printer, so I get their file specs before I
start the layout.
</task>
<requirements>
- Ask for: bleed and safe zone; minimum effective resolution; color
profile, or whether they prefer RGB and convert; preferred PDF/X
version; rich black recipe and total ink limit; proofs (hard or
PDF) and their cost; file delivery method and deadline.
- Skip anything their guidelines above already answer.
- Under 150 words, numbered questions, polite and specific.
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where support is missing; state no price or date I didn't
give.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Every question can be answered with a value or a yes or no.
- The job details are complete enough for them to quote.
- Nothing asks what their guidelines already say.
</quality_bar>
<output_format>
A subject line, then the email body in plain text.
</output_format>
<process>
End with a list of the assumptions you made.
</process>Poster critique
Stage: review · Use when: you have a draft or a proof and want a designer's review of the design (stage 4 of the chain also checks the print spec).
Quick
Critique this poster [paste a description, or attach it]: goal [e.g.
"scan to reserve"], seen from [distance, e.g. 3 m]. First say what
you see 1st, 2nd and 3rd in 3 seconds; then review hierarchy, grid,
type, color and negative space. Give the 5 changes with the biggest
effect as concrete instructions, and don't redesign it.Full
<materials>
[attach the draft or proof, or describe it: size, every element,
its position, size and colors]
[paste the brief]
</materials>
<context>
Goal: [the action, e.g. scan the QR code to reserve].
Seen: [where and from how far, e.g. a crowded board, 2–3 m].
Stage: [stage: concept | first layout | final proof]
</context>
<examples>
A poster I'd like this to be as good as: [one, with what it does
well]
</examples>
<task>
Critique the design, so I know what to fix before the next round.
</task>
<requirements>
- Start with a 3-second read from the stated distance: what is seen
1st, 2nd and 3rd, and what the designer wanted seen that was
missed.
- Then: hierarchy and the squint test; CRAP (contrast, repetition,
alignment, proximity); the grid; type (faces, sizes, spacing,
legibility at the distance); color (contrast, grayscale, CMYK
risk); negative space; the call to action.
- Don't redesign it or rewrite the copy. Don't praise it.
- Never invent numbers, names, quotes or sources: write [NEEDS
SOURCE] where support is missing.
</requirements>
<quality_bar>
Before answering, check your draft against:
- Each problem names the element and where it is.
- Each fix is a concrete instruction ("move X to Y", "set Z at
120 pt").
- Fixes are ranked by their effect on the 3-second read.
</quality_bar>
<output_format>
The 3-second read in 3 lines, a table (principle | issue | fix),
then the top 5 fixes ranked.
</output_format>
<process>
End with a list of the assumptions you made.
</process>Worked example
Weak prompt
Make a cool poster for our jazz night with a saxophone. Make it look
professional and print quality.What's wrong
- No size, ratio or print method, so an image tool returns a square, screen-resolution picture.
- "Print quality" means nothing to an image generator: bleed, CMYK and PDF/X are file settings, not a look.
- No copy and no hierarchy, so the generator invents garbled text.
- No audience, viewing distance or style, so you get generic stock-style art.
- One tool is asked to do the design and the art, and no space is left for the type.
Strong prompt
The design goes to a chat assistant, which writes the image prompt for the generator as part of its answer.
<materials>
Copy: "Late Blue: Jazz Night" · Fri 14 March, 20:00 · Union Basement
· Free entry · Lineup: [paste the 3 names] · "Scan to reserve" with a
QR code · sponsor logo.
Printer's specs: 3 mm bleed, PDF/X-4, ISO Coated v2 (FOGRA39),
300 PPI at final size.
</materials>
<context>
A free student jazz night. Goal: people scan the QR code to reserve.
It hangs on busy corridor boards and is read from 2–3 m. A2 portrait,
digital print on 170 gsm silk, 40 copies. No brand rules except the
sponsor logo, bottom right.
</context>
<examples>
Like: 1960s jazz record covers: huge bold type, one photo, two colors
plus black.
Don't like: club flyers with five fonts and a photo behind every word.
</examples>
<task>
Write the copy hierarchy, a layout spec and a CMYK palette for the
poster, plus an image-generator prompt for the art, so I can build it
in InDesign and send it to the printer.
</task>
<requirements>
- Three levels of hierarchy and one focal point.
- A 6-column grid on the A2 trim (420 × 594 mm) with 3 mm bleed and a
10 mm safe zone; positions in mm from the trim.
- At most two typefaces; sizes in pt from a type scale; each level
checked against 2–3 m (about 25 mm of letter height per 3 m).
- Palette in CMYK for FOGRA39; small text in 100% K.
- The image prompt at 2:3, with no text and the top 40% empty for
the headline.
- Never invent numbers, names, quotes or sources: write [CHECK WITH
PRINTER] where only the printer can confirm a value.
</requirements>
<quality_bar>
Before answering, check your draft against:
- The squint test: the headline leads, then date and place.
- Date and place read from 3 m; details read up close.
- Nothing important within 10 mm of the trim.
</quality_bar>
<output_format>
A hierarchy table, a layout table (element | columns | position in mm
| type), a palette table, and the image prompt in a text block.
</output_format>
<process>
Ask me up to 3 questions first if anything missing would change your
answer; otherwise go on. End with a list of the assumptions you made.
</process>What came back
The reply, illustrative, abridged:
Hierarchy
1 LATE BLUE / JAZZ NIGHT
2 Fri 14 March · 20:00 · Union Basement · Free entry
3 Lineup · "Scan to reserve" + QR code · sponsor logo
Grid: 6 columns, 10 mm gutters, 20 mm margins, 28 mm at the bottom.
Element Columns Position from top-left of trim Type
Image 1–6 full bleed -
Headline 1–6 top at 40 mm, two lines Anton 220 pt, leading 0.9
Cream band 1–6 460 mm to the bottom bleed -
Date/place 1–4 470 mm, two lines Inter Bold 36 pt
Lineup 1–3 520 mm Inter Regular 18 pt
QR code 5–6 505 mm, 32 × 32 mm -
Sponsor 5–6 545 mm -
Palette (FOGRA39)
Deep blue C100 M80 Y20 K30 image background
Warm orange C0 M55 Y90 K0 headline
Cream C0 M4 Y12 K0 bottom band
Text K100 everything on the cream band
Image prompt: Close-up of a saxophone player's hands and the bell of
the saxophone, dramatic side light, flat two-color screen-print look
in deep blue and warm orange on black, grainy texture. The subject
sits between 40% and 75% of the height, right of center; the top 40%
and the bottom 25% are plain dark blue. No text, no letters, no
logos. Aspect ratio 2:3.
Then crop about 6% off the height for A2, extend 3 mm for bleed, and
upscale to 5031 × 7087 px.
All sizes checked against the 2–3 m viewing distance.
Assumptions: Anton and Inter (both free) are acceptable; the QR code
links to the reservation page.How an expert judged it
Checked against the prompt's <quality_bar>:
- Squint test: pass. One focal point (the headline over the empty blue), then the cream band, then the details.
- Safe zone and bleed: pass. Nothing sits within 20 mm of the trim; the art runs into the bleed; the crop, bleed and upscale numbers are right for A2.
- Color: pass. Small text is 100% K on a light band, and the deep blue's total ink is 230%, under a typical limit.
- Viewing distance: fail, despite "all sizes checked". Date and place are at 36 pt. Cap height is roughly 0.7 of the point size, so 36 pt × 0.353 mm × 0.7 ≈ 9 mm, which reads from about 1 m, not 3 m. The brief put them in the 3 m check, so they need roughly 70 pt for 2 m and 100 pt for 3 m (≈ 18–25 mm caps); 36 pt is even below the 40–72 pt the size table gives for A2 secondary text. As specified, the date only works for someone already standing at the poster.
The follow-up
Date and place are at 36 pt: about 9 mm of cap height, readable from
about 1 m, but the brief needs them at 3 m. Recompute every level from
cap height (point size × 0.353 mm × the face's cap-height ratio)
against 25 mm per 3 m and show the arithmetic. Re-flow date and place
on the grid so they fit at the new size (more lines or more columns),
keep the headline at least 2× the new secondary size, and change
nothing else.Judge the output
The same checks are the ones in every <quality_bar> on this sheet. Use the table to judge any answer, and paste it
into the review stage when a model should score a draft for you in a fresh chat.
| Check | Good | Red flag |
|---|---|---|
| Hierarchy | clear 1st, 2nd, 3rd read; survives a squint test | everything the same size; two focal points |
| Message | the one message is readable from the intended distance | key info in small type, or lost in the image |
| Grid and alignment | elements share edges and axes | items floating at random positions |
| Type | at most two faces, strong size contrast, kerned headline | five fonts, default spacing, all-caps script |
| Negative space | breathing room around the focal point | every corner filled |
| Color | enough contrast; CMYK-safe; works in grayscale | neon RGB colors that will print dull; low-contrast text on photos |
| Image | no artifacts at 100%; hands, eyes and edges clean | extra fingers, melted objects, stray letters, upscaler mush |
| Specs | right trim, bleed, safe zone, effective PPI, profile, PDF/X | art stopping at the trim; text in the bleed; 72 PPI images |
| Rights | tool terms allow the use; no trademarks or lookalikes | logos, characters or a living artist's signature style |
Pitfalls
| Pitfall | Why it hurts | Instead |
|---|---|---|
| Letting the image generator set the text | misspellings, uneditable, soft edges | generate without text; set type yourself |
| Generating at the wrong ratio | cropping cuts the subject or leaves gaps | choose the closest ratio, then extend or crop |
| Upscaling one big jump | invented textures, plastic skin, mush | upscale in 2× steps and inspect each |
| Designing on screen only | RGB colors print duller; small text looks bigger on screen | soft-proof, print a scaled test, view from the real distance |
| Forgetting bleed and safe zone | white slivers at the edge; text cut off | set up the document with bleed before designing |
| 100% K for big black areas | prints as dark gray | rich black for areas; 100% K for text |
| Rich black for small text | misregistration makes it fuzzy | 100% K only for text and thin lines |
| Too many typefaces and effects | looks amateur and noisy | two faces, hierarchy from size and weight |
| Ignoring the printer's specs | rejected files, reprints | ask for their spec sheet and profile first |
| Assuming commercial rights | legal and platform risk | read the tool's terms; avoid others' IP |
References
- ISO 216 (opens in a new tab): the standard behind the A and B paper sizes
- A paper sizes (opens in a new tab): A-series dimensions in mm and inches
- Preflight files before handoff (opens in a new tab): Adobe InDesign's preflight and packaging guide
- Ghent Workgroup (opens in a new tab): industry PDF/X-based specifications for print files
- European Color Initiative downloads (opens in a new tab): free ICC profiles such as PSO Coated v3
- GRACoL (opens in a new tab): Idealliance's US printing specification and characterization data
- International Color Consortium (opens in a new tab): ICC profiles and color management
- Pantone (opens in a new tab): spot-color guides and the Pantone Matching System
- Butterick's Practical Typography (opens in a new tab): concise rules for type, including letterspacing (opens in a new tab) and line spacing (opens in a new tab)
- Google Fonts Knowledge (opens in a new tab): free lessons on typography basics and choosing type
- Copyright and Artificial Intelligence (opens in a new tab): the US Copyright Office's reports on AI and copyrightability
- Ideogram docs (opens in a new tab): an image generator known for rendering text
- Recraft (opens in a new tab): image generator with vector (SVG) output
- Vectorizer.AI (opens in a new tab): raster-to-vector tracing
- Upscayl (opens in a new tab): free open-source AI image upscaler