10 Prompts for Product Marketing with FLUX.2 Klein 4B - Wiro AI

FLUX.2 Klein 4B looks like the right kind of model for fast marketing assets: labels, badges, UI mockups, and other text-heavy images.

This post runs 10 prompts through FLUX.2 Klein Base 4B and shows the raw outputs. The focus stays on readable text and layout, not just pretty pictures.

Test setup

Model: black-forest-labs/flux.2-klein-base-4b
Resolution: 1024×1024
Steps: 35
Scale: 4
Samples: 1 per prompt
Seeds: 401 to 410

Quick takeaways

Short uppercase text works best. It tends to stay sharp and readable.
Longer words and multi-line layouts can drift. Misspellings show up (LATANTY, FINNAL, BRUNGH, COFEE).
If a prompt needs exact text, plan for a few retries or keep text short.

The 10 prompts (with outputs)

1) Product label on a mug

AI-generated studio product photo of a white mug with KLEIN 4B and FAST IMAGE API text — Prompt: Photoreal studio product photo of a matte white ceramic coffee mug on a concrete pedestal. Add clear legible text on the mug: KLEIN 4B. Smaller line below: FAST IMAGE API. Softbox lighting, shallow depth of field, clean beige background.

Clean product shot, and the text lands exactly. This style (simple background + short block of uppercase text) fits the model well.

2) Mobile onboarding UI

AI-generated isometric phone UI mockup with BUILD IN REAL TIME headline — Prompt: Isometric UI mockup of a mobile app onboarding screen. Clean modern flat design. Pastel gradient background. Big title text: BUILD IN REAL TIME. Button text: GET STARTED. Minimal icons and charts. High readability.

The big headline reads well. Smaller UI text turns into typical placeholder blur. For real app screens, keep the model focused on layout, then add exact UI copy later.

3) Latency infographic poster

AI-generated minimal latency infographic poster with misspelled header text — Prompt: Minimal infographic poster on a white background. Header text: LATENCY. Three rows with crisp text: 0.3s REAL TIME, 3s PREVIEW, 30s FINAL. Flat design, perfect typography, clean alignment.

Layout looks right, but exact spelling slips (LATANTY, FINNAL). This pattern shows up a lot: the model nails alignment, then fumbles one or two characters.

4) Neon storefront sign

AI-generated night bakery storefront with neon FRISCHE BROT sign — Prompt: Cinematic night photo of a small bakery storefront in Berlin. Wet pavement reflections. Neon sign with clear German text: FRISCHE BROT. Smaller window sign text: HEUTE OFFEN. 35mm film look, high detail.

The main sign reads correctly. The smaller sign drops a letter (HEUT OFFEN). Tiny text stays the hardest part.

5) Skincare bottle label

AI-generated skincare bottle product shot with CLEAR GEL label — Prompt: Photoreal studio shot of a clear skincare bottle with a matte label. Label text: CLEAR GEL. Small line: 2 PERCENT NIACINAMIDE. Bottom line: 50 ML. Clean minimal design, soft lighting, shallow depth of field.

Three lines of label text come out readable. This kind of label prompt works well when the copy stays short and high-contrast.

6) Conference badge

AI-generated conference badge with WIRO DEV and BUILD FAST text — Prompt: Photoreal conference badge on a lanyard on a desk. Badge has crisp text: WIRO DEV DAY. Smaller line: BUILD FAST. Minimal design, soft shadows, realistic materials, product photo style.

It keeps the main idea but shortens the big line to WIRO DEV. If the badge needs an exact event name, prompt retries help.

7) Chalkboard cafe menu

AI-generated chalkboard brunch menu with misspellings in the title and items — Prompt: Realistic chalkboard cafe menu. Title text: BRUNCH MENU. Six items with prices, all legible: EGGS 6, PANCAKES 8, AVOCADO TOAST 9, OATMEAL 5, COFFEE 3, TEA 2. Handwritten chalk style, soft vignette.

Chalk handwriting looks convincing, but exact spelling drifts (BRUNGH, COFEE). Handwritten styles trade accuracy for vibe.

8) Vintage travel postcard

AI-generated Lisbon tram postcard with HELLO FROM LISBON and TRAM 28 text — Prompt: Vintage travel postcard on a wooden table. Illustration of a yellow tram in a steep street. Postcard headline text: HELLO FROM LISBON. Small line: TRAM 28. Warm colors, paper texture, crisp lettering.

The postcard format gives the model strong structure. Both lines read correctly, and the paper texture sells the print feel.

9) Movie title card

AI-generated movie title card with misspelled SIGNAL LOST text — Prompt: Minimal movie title card. Dark background with subtle grain. Large centered title text: SIGNAL LOST. Subtitle text below: A SMALL MACHINE STORY. Clean serif typography, high contrast, no extra text.

High contrast helps. Still, the main title shifts (SIGIAL LOST). Short words can still break when the model tries to stylize typography.

10) Sticky note handwriting

AI-generated laptop desk scene with sticky note that says SHIP IT TODAY and FIX BUG 192 — Prompt: Photoreal office desk scene. Open laptop on a wooden desk. A yellow sticky note on the laptop bezel with clear handwritten text: SHIP IT TODAY. Small line below: FIX BUG 142. Soft morning window light.

The note reads cleanly, but the bug number changes (192 vs 142). Numbers can wander even when letters hold up.

Tips to get more reliable text

Keep copy short. One or two words per line works better than sentences.
Use uppercase for labels and badges. It tends to stay clearer.
Ask for high contrast: dark text on light surfaces, or bright neon on dark scenes.
If exact spelling matters, generate the background and add text in a design tool.

Try the model here: FLUX.2 Klein Base 4B on Wiro.