Nano Banana 2 Infographic
Create sleek, rich, non-noisy infographic prompts and review sets for Gemini image generation.
This skill uses Nano Banana 2 only. For API calls, use the live callable model ID rather than assuming the public marketing name is the exact endpoint name.
Decision Tree
What do you need to do?
- The brief is incomplete or fuzzy Ask only for the missing essentials: topic, audience/context, must-include facts, and brand or style constraints.
- The user wants an infographic now Prepare four review variants by default at
16:9unless the user specified another ratio. Readreferences/patterns.md. - The user wants live Gemini renders or proof that the prompt works Read
references/configuration.md, then runscripts/probe_gemini_image_api.py. - The user wants exact API syntax, model IDs, or request fields Read
references/api.md. - The result looks noisy, text-heavy, or poster-like Read
references/gotchas.md, simplify the composition, and regenerate.
Default Operating Mode
- Offer four distinct variants by default unless the user explicitly asks for fewer.
- Default aspect ratio to
16:9. - Use Nano Banana 2 only. Do not fall back to older image models unless the user explicitly asks.
- Render the default pack concurrently when you need live outputs fast.
- Use separate render passes for the variants instead of trusting one request to return the exact number of images requested.
- Keep visible text short: title up to 5 words, labels 1-3 words, no paragraphs in the image.
- Prefer editorial restraint over maximal detail. If a choice would add noise, cut it.
Intake Questions
Ask these only when they are not already answered:
| Missing | Ask |
|---|---|
| Topic or claim | "What is the infographic about, in one sentence?" |
| Audience or channel | "Where will this live: blog post, deck, report, keynote, or something else?" |
| Facts or sections | "Which numbers, claims, or sections must appear?" |
| Style boundaries | "Any brand colours, must-avoid looks, or reference tone?" |
If the user already gave the essentials, do not re-interview them. Build the variant pack immediately.
Quick Reference
| Need | Do | Output |
|---|---|---|
| Fast prompt pack | Run scripts/build_variant_pack.py with a brief JSON | four prompt variants plus a markdown review sheet |
| Fast parallel render | Run scripts/render_variant_pack.py on variant-pack.json | all variants rendered concurrently plus a batch manifest |
| Live render proof | Run scripts/probe_gemini_image_api.py on one prompt | saved response JSON and local image files |
| Default professional set | Use Executive Snapshot, Editorial Column, Decision Board, and Insight Ribbon | four reviewable directions |
| Noise reduction | Remove extra panels, colors, and prose before re-rendering | cleaner second pass |
Default Variant Quartet
| Variant | Best For | Direction |
|---|---|---|
| Executive Snapshot | C-suite slides, board pre-reads, strategic summaries | one dominant claim or number with 3-4 disciplined support blocks |
| Editorial Column | Blog posts, reports, explainers | tall stacked panels with generous whitespace and thin dividers |
| Decision Board | trade-offs, frameworks, comparisons | modular grid or side-by-side layout with equal visual weight |
| Insight Ribbon | keynote hero slides, opener visuals, and wide summaries | one horizontal narrative band with evenly spaced support modules |
Read references/patterns.md for the exact prompt shape and regeneration ladder.
Rendering Rules
- Say
16:9explicitly unless the user asked for another ratio. - Ask for a white or near-white base, restrained accents, and flat editorial graphics.
- Keep to 2-3 accent colours plus gray or white.
- Use one visual idea per image. Do not combine process, comparison, glossary, and hero illustration in the same render.
- Put the long explanation outside the image. Generate the copy first, then render only the short text that must appear.
Gotchas
- Asking for a "detailed infographic" usually increases clutter rather than clarity. Ask for hierarchy, whitespace, and restraint instead.
- Google documents that the model might not create the exact number of images requested. Treat the four default variants as four deliberate passes.
- Google also documents that text generation works best when the text is decided first and then rendered into the image. Do not improvise long copy inside the image prompt.
- If the image looks like a poster, reduce the number of panels, colors, and icon families before changing everything else.
- When the user needs dense quantitative fidelity, hand-built charts or vector layouts may be a better fit than Gemini image generation.
Reading Guide
| Task | Read |
|---|---|
| Model IDs, request fields, aspect ratios, response shape | references/api.md |
| Variant design, question flow, prompt formula, iteration ladder | references/patterns.md |
| Environment setup, scripts, and live verification commands | references/configuration.md |
| Noise, text, language, and retry pitfalls | references/gotchas.md |