Gadgetify shares a GPT Image 2 prompt for knolling on a 2×2 grid with themed silhouettes

On June 19, Gadgetify (@Gdgtify) posted to X a knolling experiment — overhead photography of ordered objects aligned at 90° — generated with AI. He tried the same template in Nano Banana and ChatGPT's GPT Image 2: “he tried, but it didn't turn out that well”; with OpenAI, “kind of nailed it.” The prompt asks for a 2×2 grid with four themes inferred by the AI, each with 20–30 objects whose collective outline forms a silhouette (wave, chef's hat, Mount Fuji…).

AI Prompt · How to do it

Example posted by Gadgetify on June 19, 2026. @Gdgtify

Steps

1. Open chatgpt.com with GPT Image 2 active.
2. Copy the prompt below or replace [Theme] and [Master Shape] (e.g. "a Surfer" + "a wave").
3. For a single panel, remove “2x2 grid” and order a single theme.
4. If there is overlap or clutter, reinforce the Negative block at the end.

Prompt to copy

2x2 grid, do this for 4 AI inferred themes and shapes Anchor: A perfectly organized top-down knolling flat-lay of every object associated with [Theme (e.g. a Surfer / a Baker / Tokyo / the 1980s)] ::4 Macro Shape: Objects arranged so their collective outline forms the silhouette of [Master Shape (e.g. a wave / a chef's hat / Mount Fuji)] — order within, meaning at the edge. Infer 20–30 fitting items ::5 Composition: Strict 90° grid alignment, even spacing, generous margin ::3 Material Physics: Real objects, soft studio shadows, consistent matte finish ::2 Illumination: Flat overhead softbox, no glare ::2 Render Stack: Knolling photography, organized-neatly aesthetic, product flat-lay ::1 Negative: [tilted objects, overlap, mess, motion blur, duplicate items, off-theme objects] :: -1

Syntax with ::n weights (Midjourney style). In English according to the original post. Results vary by ChatGPT plan and region.

What makes this prompt special

It doesn't just ask for "ordered objects": it combines knolling (strict alignment, studio shadows, organized-neatly aesthetics) with a macroform - the objects, together, must draw the silhouette of a wave, a hat or a mountain. The weights ::4, ::5, etc. They prioritize thematic anchor, number of items and composition over lighting or finish.

The Negative block with negative weight penalizes tilted objects, overlaps, clutter, and duplicates — typical flaws in AI-generated knolling. Gadgetify contrasts it with Nano Banana in the same thread, which positions GPT Image 2 as a better option for this type of complex flat-lay.

Theme and form ideas

• Surfer → silhouette of wave
• Baker → chef's hat
• Tokyo → Mount Fuji
• 1980s → cassette or arcade cabinet

Before publishing

Visible markings on generated objects may be random; do not use results with recognizable logos in commercial campaigns without review. Respect OpenAI policies for image generation.