Marketing mockups with real text are the highest-value use of AI image generation. Get it right and you have a hero image for a landing page, a social ad creative, or a packaging visualization that ships the same day. Get it wrong and the headline is illegible, the brand name is misspelled, or the screen text is gibberish.
This is the production guide: the 6 prompt patterns for marketing mockups, the model pick by mockup type, the typography and layout rules, the brand-safety constraints, and the workflow for shipping marketing visuals that hold up at any size.
Why marketing mockups with real text are a different problem
Marketing mockups sit at the intersection of three hard problems: text rendering, brand consistency, and product visualization. The mockup has to render the actual text (the headline, the brand name, the screen UI), preserve the brand's visual identity (colors, fonts, layout), and look like a real product visualization (lighting, scale, context).
Three things make this hard:
- Text rendering at small sizes. Marketing mockups have text at 6-18pt (button labels, navigation, captions). The model must render small text legibly without typos.
- Brand consistency across the mockup. The headline, body, button, and brand mark must look like they came from the same design system. Same font, same color palette, same spacing.
- Real-product context. The mockup has to look like a real product (app screen, packaging, billboard) in a real context (in a hand, on a shelf, in a city). The visual has to convince.
The fix is to pick the right model for the mockup type, structure the prompt with the text in quotes, and verify the output before publishing.
The model pick by marketing mockup type
| Mockup type | Best model | Why |
|---|---|---|
| App screen / UI mockup | GPT Image 2 | Strongest text-in-image, best at UI layouts |
| Landing page hero | GPT Image 2 | Best at multi-element layouts with text |
| Social media ad creative | GPT Image 2 or Ideogram V3/V4 | Strong at branded layouts |
| Packaging mockup (box, bottle, bag) | GPT Image 2 or Nano Banana 2 | Strong at product visualization + text |
| Billboard / out-of-home ad | GPT Image 2 or Ideogram V3/V4 | Strong at large-format text |
| Email header graphic | GPT Image 2 | Best at text-heavy layouts |
| Webinar / event promo | GPT Image 2 or Ideogram V3/V4 | Strong at branded event graphics |
| T-shirt / merch mockup | GPT Image 2 or Ideogram V3/V4 | Strong at product + text |
| Book cover / podcast cover | GPT Image 2 or Ideogram V3/V4 | Strong at typography-led design |
| Presentation slide / pitch deck | GPT Image 2 | Best at structured layouts |
Key takeaways:
- GPT Image 2 wins on most marketing mockups. Strongest text-in-image accuracy (99% on short text), best at dense multi-element layouts.
- Ideogram V3/V4 is the strong alternative for branding-led mockups (logos, posters, social cards).
- Nano Banana 2 wins on packaging and product visualization when you need photoreal product + text.
The 6 prompt patterns for marketing mockups
Pattern 1: App screen / UI mockup (mobile app, dashboard, web UI)
The "show the product interface" mockup. Used for product launches, app store screenshots, investor decks, marketing site hero images.
The prompt structure (GPT Image 2, optimized for UI text):
A vertical mobile onboarding screen for a fictional app called
"NESTING". The screen is a phone mockup (iPhone 15 Pro) viewed
straight-on, on a soft warm cream background.
Screen content (top to bottom):
- Status bar: time "9:41", signal bars, battery icon
- Logo: "N" mark in coral (#E07A5F), centered, 24pt
- Headline: "WELCOME TO NESTING" in bold rounded sans-serif,
centered, 28pt, color: charcoal (#1F2937)
- Subhead: "A quieter way to gather people around a table" in
regular rounded sans-serif, centered, 16pt, color: medium gray
(#6B7280)
- Primary button: "Get started" in white on coral (#E07A5F),
centered, 18pt bold, full-width with rounded corners (8px)
- Secondary button: "I already have an account" in coral
(#E07A5F) on transparent, centered, 14pt regular
- Bottom illustration: a small line drawing of three plates and
two wine glasses, centered, 80x60px
Constraints: photorealistic phone frame with realistic shadows,
no watermark, no real app branding, all text must be exactly
as written, no fake testimonials, no fake user counts.
On GPT Image 2 specifically: the strongest model for UI mockups. Specify the screen content top-to-bottom, the typography, the colors in hex, and the brand-safety rules.
On Seedream 5.0 Lite specifically: the reasoning model is good at "explain the screen" — useful for app flows.
The brand-safety rules:
- "no real app branding" (if it's a fictional app)
- "no fake user counts"
- "no fake testimonials"
- "all text must be exactly as written"
- "no invented features"
Pattern 2: Landing page hero (web hero, marketing site, blog header)
The "show the product on the website" mockup. Used for marketing sites, blog headers, feature announcement graphics.
The prompt structure (GPT Image 2, optimized for hero layouts):
A modern marketing website hero section. Layout: a 16:9
landscape, with the hero on the left half and a product
mockup on the right half.
Left side (text content, left-aligned):
- Eyebrow text: "NEW • 2026" in small uppercase, letter-spaced
100 units, color: brand primary (#1E40AF), 12pt
- Headline: "Ship 10x faster with the prompt library that
actually works" in bold sans-serif, 48pt, color: charcoal
(#1F2937)
- Subhead: "Browse 1,200+ tested prompts for Nano Banana 2,
GPT Image 2, and Seedream. Copy, paste, ship." in regular
sans-serif, 18pt, color: medium gray (#6B7280)
- Primary CTA: "Start free trial" in white on brand primary
(#1E40AF), bold 16pt, rounded corners, padding 12px 24px
- Secondary CTA: "Watch demo" in brand primary on transparent,
16pt
Right side (visual): a stylized product mockup showing a
browser window with a prompt library interface. The interface
is abstract — cards, tags, and a search bar — but no fake brand
names.
Background: subtle gradient from white at the top to light
blue (#EFF6FF) at the bottom. Modern minimal style, photoreal,
no watermark, no fake testimonial quotes, no fake customer
logos, all text must be exactly as written.
On GPT Image 2 specifically: the strongest model for landing page heroes. Specify the layout (left/right), the typography, the colors, the brand-safety rules.
On Ideogram V3/V4 specifically: strong on branding-led hero sections.
The brand-safety rules:
- "no fake testimonial quotes"
- "no fake customer logos"
- "all text must be exactly as written"
- "no fake product claims"
- "no fake company names"
Pattern 3: Social media ad creative (Instagram, TikTok, Facebook)
The "make people stop scrolling" mockup. Used for paid social, organic posts, story ads.
The prompt structure (GPT Image 2 or Ideogram V3):
A vertical Instagram Story ad for a coffee brand.
Layout (top to bottom):
- Top zone: a high-resolution photograph of a latte being
poured into a cream ceramic mug, with steam rising, soft
morning light
- Middle zone (overlaid on the photo): brand mark "ROASTERY"
in cream serif, top-left
- Headline (centered over the photo): "MORNINGS, SLOWER" in
cream bold sans-serif, 48pt
- Subhead (centered below headline): "small batch, slow
roasted, delivered weekly" in cream regular sans-serif, 16pt
- Bottom zone: a CTA sticker "ORDER NOW →" in white on brand
brown (#3E2723), centered, with a subtle drop shadow
Constraints: 9:16 vertical, no fake URLs, no fake pricing,
no fake customer reviews, all text must be exactly as
written, no watermark, no fake brand partnerships.
On Ideogram V3/V4 specifically: strong on branding-led social cards. Specify the brand colors in hex.
On GPT Image 2 specifically: strong on photo + text overlay. Specify the photo, the overlay text, the colors, the CTA.
The brand-safety rules:
- "no fake URLs"
- "no fake pricing"
- "no fake customer reviews"
- "all text must be exactly as written"
- "no fake brand partnerships"
- "no fake influencer endorsements"
Pattern 4: Packaging mockup (box, bottle, bag, label)
The "show the product packaging" mockup. Used for product launches, ecommerce hero images, investor decks, brand guidelines.
The prompt structure (GPT Image 2 or Nano Banana 2):
A photorealistic product photo of a coffee bag standing
upright on a wooden surface, with a soft natural light from
the upper-left.
The bag is a kraft paper bag with a matte finish, 12oz size,
with a flat bottom and a tin tie at the top. The label is a
cream rectangle on the front, with the following content:
- Brand mark: "ROASTERY" wordmark in serif, top of label,
color: brand brown (#3E2723)
- Origin line: "ETHIOPIA • YIRGACHEFFE" in small uppercase
sans-serif, letter-spaced 100 units
- Coffee name: "Konga" in italic serif, large
- Tasting notes: "jasmine, bergamot, honey" in regular
sans-serif
- Weight: "12 oz / 340g" in small sans-serif
- Roast date line: "ROASTED: 2026-06-15" in small sans-serif
Background: a soft blurred kitchen scene, warm color
temperature (4500K). The bag fills 70% of the frame, in
sharp focus with a slight depth of field falloff on the
background.
Photorealistic, 1080x1080, no watermark, no fake brand
names, no fake awards, all text must be exactly as written.
On Nano Banana 2 specifically: strong at photoreal product + text. Use multi-image references for product + label.
On GPT Image 2 specifically: strong at packaging + text. Specify the label content top-to-bottom, the typography, the colors.
On Seedream 5.0 Lite specifically: good at reasoning over packaging requirements ("the label must look like a real coffee label, not a fake").
The brand-safety rules:
- "no fake brand names"
- "no fake awards"
- "no fake certifications"
- "all text must be exactly as written"
- "no fake origin claims"
- "no fake roast date" (or use the actual date)
Pattern 5: Billboard / out-of-home ad (large format)
The "show the brand at scale" mockup. Used for OOH campaigns, transit ads, billboard concepts.
The prompt structure (GPT Image 2 or Ideogram V3):
A photorealistic billboard mockup on a city street. The
billboard is 14x48 feet, mounted on a steel frame, viewed
from the sidewalk at a 3/4 angle. The setting is a busy
downtown street at golden hour, with soft warm light on
the billboard.
The billboard ad has the following content (left to right):
- Brand mark: "ACME" wordmark in bold sans-serif, top-left
of the billboard, color: white on brand primary (#1E40AF)
background
- Headline: "BUILD SOMETHING THAT MATTERS" in bold white
sans-serif, large (60pt equivalent)
- Visual: a stylized geometric illustration of a building
rising from a foundation
- Footer: "acme.com" in regular white sans-serif, bottom-right
Background: golden hour city street, slightly out of focus.
The billboard is in sharp focus, color-accurate, and the
text is legible from the sidewalk.
Photorealistic, 16:9, no watermark, no fake URLs, all text
must be exactly as written.
On GPT Image 2 specifically: strong at large-format text. Specify the size equivalent, the typography, the color contrast.
On Ideogram V3/V4 specifically: strong on branding-led billboards.
The brand-safety rules:
- "no fake URLs"
- "all text must be exactly as written"
- "no fake claims"
- "color must be brand-accurate"
Pattern 6: T-shirt / merch mockup (apparel, accessories)
The "show the merch" mockup. Used for product launches, ecommerce, brand merchandise.
The prompt structure (GPT Image 2 or Ideogram V3):
A photorealistic t-shirt mockup. A heather gray cotton t-shirt
on a male model, 30, with a relaxed fit, viewed from the
front. The t-shirt is the focal point, in sharp focus, with
a slight depth of field falloff on the background.
The t-shirt design (centered on the chest):
- Top: "ACME • 2026" in small sans-serif, color: charcoal
(#1F2937), letter-spaced
- Middle: a stylized "A" mark in brand primary (#1E40AF),
large
- Bottom: "BUILT TO LAST" in small sans-serif, color:
charcoal, letter-spaced
Background: a soft blurred urban setting, warm color
temperature (4500K). Photorealistic, 1080x1080, no watermark,
no fake brand marks, all text must be exactly as written.
On GPT Image 2 specifically: strong at apparel + text. Specify the design top-to-bottom, the colors, the typography.
On Ideogram V3/V4 specifically: strong on branding-led merch.
The brand-safety rules:
- "no fake brand marks"
- "all text must be exactly as written"
- "no fake slogans"
- "no fake quotes"
The brand-safety constraints (universal)
Across all 6 patterns, these are the brand-safety rules every marketing mockup prompt should include:
- "All text must be exactly as written" — prevents the model from paraphrasing
- "No fake URLs" — prevents invented website addresses
- "No fake testimonials / quotes" — prevents fake customer voices
- "No fake customer logos" — prevents invented brand partnerships
- "No fake statistics" — prevents hallucinated numbers
- "No fake user counts" — prevents "10,000+ users" type claims
- "No fake pricing" — prevents invented dollar amounts
- "No fake brand names" — prevents invented competitors or partners
- "No fake awards / certifications" — prevents invented recognition
- "No real brand logos unless licensed" — prevents copyright issues
The list is long. The point is: AI models will fill any void in the prompt with plausible-looking fake content. The brand-safety rules are the walls that keep the fake content out.
The typography rules for marketing mockups
- Use one font family. Helvetica, Inter, Roboto, IBM Plex Sans. Two weights max (Regular, Bold).
- Establish a type scale. 10pt / 12pt / 14pt / 18pt / 24pt / 32pt / 48pt / 60pt. Pick 3-4 per mockup.
- The headline is the largest type. 32-60pt. The body is 14-18pt. The footer is 10-12pt.
- Color the type with intent. Headline in brand primary or white-on-photo. Body in charcoal or medium gray. Footer in medium gray.
- Specify contrast. White on dark, dark on light. The model needs to know the contrast direction.
- Use the model's typography strength. GPT Image 2 holds small text best. NB2 is good at 14pt+. Seedream is good at labels.
The marketing-mockup pre-flight checklist
Before you ship a marketing mockup:
- All text is verified. Re-read every word. AI output has typos.
- All URLs are real or removed. No fake URLs.
- All statistics are verified. No fake numbers.
- The brand colors are accurate. Verify hex codes match the brand guide.
- The brand-safety rules are applied. 10 rules above.
- The font is consistent. One font family, 2-3 weights.
- The text is legible at target size. Open at the actual size (1080x1080, 1080x1920, etc.).
- The visual context is appropriate. App screen on a real device, packaging on a real surface, billboard on a real street.
- The post-processing pipeline is applied. Color correction, color grading, sharpening, grain.
- The provenance is documented. Model used, prompt, date, disclosure (per the safety post).
Skip any of these and the mockup is either wrong, non-compliant, or non-brand-safe.
The iteration loop for marketing mockups
Pass 1: Generate with all text in the prompt. Every word, every URL, every stat in quotes.
Pass 2: Verify the text. Re-read the output character by character. AI output has typos.
Pass 3: Verify the brand-safety rules. No fake URLs, no fake testimonials, no fake stats. Regenerate with explicit negatives if needed.
Pass 4: Verify the visual. The mockup should look like a real product in a real context. If the lighting is off, the scale is wrong, or the context is generic, refine the prompt.
Pass 5: Apply the post-processing pipeline. Color correction, color grading, sharpening, grain.
Pass 6: Final verification at target size. Open at the actual size and verify readability. If text is illegible, regenerate at higher resolution or with larger type.
Pass 7: Verify disclosure requirements. Per the safety post, the model used, the date, and the disclosure are recorded.
The summary
Marketing mockups with real text are the highest-value use of AI image generation. The 6 patterns — app screen, landing page, social ad, packaging, billboard, merch — cover ~90% of marketing needs.
- Pick the model for the mockup type. GPT Image 2 wins on most, Ideogram V3/V4 for branding-led, Nano Banana 2 for product packaging.
- Put every word in the prompt in quotes. Literal text, not paraphrased text.
- Apply the 10 brand-safety rules. No fake URLs, no fake testimonials, no fake stats, no fake anything.
- Use the typography rules. One font family, type scale, intentional color, model-appropriate size.
- Verify the output character by character. AI output has typos. Re-read every word.
The model is not the problem. The brand safety and text accuracy are. Put the text in quotes, apply the brand-safety rules, verify the output, and marketing mockups become a reliable, brand-safe, publication-ready workflow.
