Idea to Visual

Turn one idea into one clear, simple visual that lands in a few seconds. Image models are good renderers and poor concept-makers. Given a bare prompt like "illustrate this quote", they produce average, literal images. The value is in the brainstorm before the prompt. This skill runs that brainstorm, filters the concepts hard, and only then writes the prompt and generates the image.

Core principles (apply to every step)

1. One idea per visual. Pick a single takeaway. This is not an infographic.
2. Clarity beats cleverness. If a clever version confuses, drop it. Test: would a stranger get it in 3–4 seconds, even if they disagree with it?
3. Novelty. Show a familiar idea from an unexpected angle, through an unexpected combination, or so simply that it feels new.
4. Resonance. Name what the audience feels, fears, thinks, or says. Relatable moments beat abstractions.
5. Amplify, don't repeat. The text and caption should add a punch, a question, or advice on top of the image, not describe it.
6. Flow. Elements and words should read in order and click together, sometimes with a rhythm or rhyme ("easy to start / hard to stop").
7. Few words, few text placements, big shapes. Generators drop small details like tiny arrows and fine labels, and extra text clutters. Keep total on-image text short.
8. Taste is the filter. Most generated concepts are mediocre. Treat them as fuel for the next step, and pull out useful fragments (a layout, a metaphor, a phrase) even when the whole concept fails.

Pick the mode

Infer the mode from the request. Ask only when the source idea itself is missing.