JSON Prompt Template — App Icon
# JSON Prompt Template — App Icon
A focused JSON template optimized specifically for app icons. Based on the generic 6-part schema from the wider Nano Banana 2 prompting guide, but trimmed to the fields that matter for icons and pre-filled with non-negotiable constants.
## Why JSON (not prose)
Nano Banana 2 produces more predictable output from structured briefs than from natural-language prompts. The structured format also makes iteration trivially — in Step 6 (refinement), the user swaps one field instead of rewriting an entire paragraph.
Never convert these prompts to prose. Always keep them as JSON.
## Mapping table — brief answer → JSON field
| Brief line | JSON field | Notes |
|---|---|---|
| Central concept (Q2) | `subject.description` | The ONE thing. Phrase as a noun with 1–2 adjectives. |
| Style family (Q7) | `subject.style` + `technical.format` | Duplicated on purpose — Nano Banana weighs both. |
| Adjectives (Q5) | `subject.mood` | Exactly 3 tokens. |
| Palette primary (Q4) | `palette.primary` | Comma-separated hex codes. |
| Palette background (Q4 + Q9) | `palette.background` | Single hex, prefixed with "solid". |
| Palette accent (Q4) | `palette.accents` | Optional highlights. |
| Shape language (Q8) | `composition.shape_language` | Verbatim from the enumerated list. |
| Competitors (Q6) | `negative_constraints` | Append as `"do not resemble: X, Y"`. |
## Non-negotiable constants
These values are always the same for app icons. Never let the user override them in the initial prompt — they encode Apple HIG-style constraints that are the difference between "AI slop" and "designed icon".
- `image_type: "app icon"`
- `composition.framing: "centered, subject fills 70% of canvas"`
- `composition.detail_level: "very low — big recognizable shapes only, must read at 48x48px"`
- `technical.resolution: "1024x1024"`
- `technical.background_type: "solid color"`
- `technical.constraints: "no text, no letters, no fine details, no realistic textures, icon-friendly, nothing outside the frame"`
## The template (fill in the placeholders)when to use it
Community prompt sourced from the open-source GitHub repo rolandtolnay/llm-toolkit (MIT). A "JSON Prompt Template — App Icon" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.
tags
image-gencommunitygeneral
source
rolandtolnay/llm-toolkit · MIT