A 2×2 ad-banner grid in the visual register of dense Japanese digital advertising — bold subjects, layered headline copy, tiny price-style labels, small badge clusters. Hand over four panel themes (travel / skincare / food / education / cosmetics / electronics / fashion / restaurant), the subject and headline copy for each, pick a primary language and an optional brand mark, and the model produces a single composed image with 4 quadrants — each its own self-contained ad with theme-correct typography, palette, and badge cluster.

What you bring

Four panels (top-left → top-right → bottom-left → bottom-right) — for each: a theme, a subject (what's depicted), and a headline (the largest text in the panel). The model treats each headline as literal copy and renders it verbatim.
A primary language — Japanese (most authentic to the original style), English, Simplified Chinese. All headline copy, panel labels, badge text, and price tags render in this language consistently.

Optionally: a brand mark that sits as a tiny restrained mark in one corner of the entire grid.

What you get back

A single composed 1:1 image with 4 distinct ad banners arranged in a 2×2 grid. Each panel has one clear hero subject, headline copy in the chosen language, and 3-5 small price-style labels / badges / chips clustered around the headline. Each panel's palette is distinct from the other three so the four quadrants don't blur together. Reads as an actual ad-agency portfolio piece — usable for multi-vertical campaign briefs, agency Behance / Dribbble posts, or X campaign-pitch images.

Three languages, no cross-leak

Each language has its own typography vocabulary and ad-vocab register — Japanese gets 期間限定 / 〜円〜 / 〜万人突破 patterns, English gets FROM $XX / NEW / LIMITED, Simplified Chinese gets ¥XX起 / 限时 / 立减. The chosen language is binding across the entire grid; mixed-script slip is forbidden, traditional Chinese is forbidden in Simplified Chinese mode, and Japanese kana is forbidden in English mode. Real-celebrity faces and real-platform logos are also forbidden — every panel is original copy.

Repeater slot — exactly 4 panels

This Look uses the repeater slot type with a min and max of 4. The form gives you four grouped sub-fields (theme / subject / headline) per panel. You can't add or remove panels — the 2×2 layout is locked. If you want fewer panels per Look, use the Brand Mascot Identity Board (6 panels of brand identity) or the Viral Case Study Page (5 stacked sections).

Cost & timing

8 credits per run. Roughly 70-80 seconds. With your 20 signup credits you get two before topping up.

Try it →

Inspired by @makaneko_AI on X — the seed prompt presenting a 2×2 grid of Japanese digital ad banners with travel / skincare / food / education themes, each with subject + headline + badge cluster. The three-language enforcement (Japanese / English / Simplified Chinese), locked 2×2 layout grid, repeater slot for the 4 panels, instruction fence, and per-language FORBIDDEN-cue blocks (no script bleed, no celebrity faces, no real-platform logos) are FluxGen's.