Generate one optimized Nanobanana Pro prompt for a blog image.
Step 1: Pick the mode
| Mode | Use when the user wants... | Goal |
|---|---|---|
| Cover | A header, hero, thumbnail, or scroll-stopper | Visual impact — abstract and attention-grabbing |
| Illustration | An in-post diagram, workflow, comparison, or concept explainer | Clarity — literal and educational |
If the user supplies a blog title for a cover, treat it as a titled cover (title becomes the focal point).
Step 2: Apply the mode spec (subagent)
Read the corresponding reference file for the chosen mode:
- Cover →
references/cover.md - Illustration →
references/illustration.md
Apply all style rules, palette constraints, and formatting guidelines from that file.
Step 3: Write the prompt
Use this structure:
Mode: [Cover / Illustration]
Focus: [One-line description of the visual's purpose]
[ASPECT] [TYPE], [STYLE DESCRIPTORS]. [BACKGROUND].
[LAYOUT — structure, placement, spacing]
[ELEMENTS — each with color, position, label, visual cue]
[RELATIONSHIPS — what connects, groups, or contrasts]
[STYLE NOTES from the mode spec above]Keep the prose tight. Every sentence should either describe something the generator needs to draw or a style rule it needs to follow.
Step 4: Review and tighten
Before returning the prompt, check it once:
- Remove repeated facts or labels.
- Labels should name entities of the diagram, not effects.
- Return only the cleaned-up prompt.