/image

description: Generate and edit images using AI. Powered by Nano Banana Pro (Google Gemini image models) with Pollinations.ai and Imagen 4 fallback. Supports text-to-image, image editing, aspect ratios, 2K/4K, and batch generation. Use when generating images, creating visuals, AI art, text-to-image, image generation, create picture, make illustration, generate photo, nano banana, edit image, batch images.
version: 1.0.0
allowed-tools: Read, Bash, Glob
context: fork

Image Generation Skill

Generate and edit images from text prompts using AI. Powered by **Nano Banana Pro** (Google Gemini image models — `gemini-3.1-flash-image-preview`, `gemini-2.5-flash-image`, `gemini-3-pro-image-preview`), with Pollinations.ai and Imagen 4 as fallbacks.

> **Note**: If the `nanobanana` skill is available in your session, this skill delegates to it automatically. Otherwise, it uses its own built-in Nano Banana implementation as fallback.

Provider Fallback Chain

Standard Mode (default — optimizes for cost)

Tier 1: Gemini Native (FREE) ─── gemini-3.1-flash-image-preview (Nano Banana 2) ──┐
        ↓ on error                                                                   │
        gemini-2.5-flash-image ─────────────────────────────────────────────────────┤
        ↓ on error                                                                   │
        gemini-3-pro-image-preview (Nano Banana Pro) ───────────────────────────────┤
        ↓ on error                                                                   │
Tier 2: Pollinations.ai (FREE, no key) ─────────────────────────────────────────────┤
        ↓ on error                                                                   │
Tier 3: Imagen 4 (PAID, billing required) ──────────────────────────────────────────┘

High-Quality Mode (`--hq` or "high quality" in prompt)

Tier 1: Imagen 4 (PAID, ~$0.04/image) ────────────────────────┐
        ↓ on error                                              │
Tier 2: gemini-3-pro-image-preview (Nano Banana Pro) ──────────┤
        ↓ on error                                              │
Tier 3: gemini-3.1-flash-image-preview (Nano Banana 2) ────────┤
        ↓ on error                                              │
Tier 4: Pollinations.ai (FREE) ────────────────────────────────┘

Supported Features

| Feature | Nano Banana 2 | Nano Banana Pro | Pollinations | Imagen 4 | |---------|:---:|:---:|:---:|:---:| | Text-to-image | ✓ | ✓ | ✓ | ✓ | | Image editing | ✓ | ✓ | — | — | | Aspect ratios | ✓ | ✓ | ✓ | ✓ | | 2K/4K output | ✓ | ✓ | — | ✓ | | Search grounding | ✓ | ✓ | — | — | | Batch generation | ✓ | ✓ | ✓ | ✓ |

**Aspect ratios supported**: `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`

Workflow

Step 0: Delegate to Nano Banana Skill (if available)

Before running the built-in fallback chain, check if the `nanobanana` skill is available in the current session. If it is, delegate the entire request to it — it provides a richer, dedicated Gemini image generation experience.

**How to check**: Look for `nanobanana` in the available skills list (system-reminder). If present:

1. Invoke `Skill({ skill: "nanobanana" })` with the user's original prompt 2. **Done** — do not continue to Step 1. The `nanobanana` skill handles everything (generation, editing, batch, aspect ratios, 2K/4K).

**When to skip and use built-in chain instead**:

• `nanobanana` skill is NOT listed in available skills
• User explicitly requests Pollinations or Imagen 4
• User says "use fallback" or "don't use nanobanana"

If `nanobanana` is not available, proceed with the built-in chain below.

Step 1: Parse User Request

Extract from the user's prompt:

• **Subject**: What to generate or edit
• **Style**: Photorealistic, illustration, pixel art, etc. (default: photorealistic)
• **Quality**: `standard` or `high`. Detect from: "high quality", "hq", "best quality", "maximum quality", "premium"
• **Resolution**: `standard`, `2K`, or `4K`. Detect from: "2K", "4K", "high-res", "high resolution"
• **Aspect ratio**: Detect from explicit mention (e.g., "16:9", "square", "portrait", "widescreen", "cinematic")
• Common aliases: square→`1:1`, portrait→`9:16`, landscape→`16:9`, widescreen→`21:9`, vertical→`9:16`
• **Input image**: If user provides an image path for editing
• **Output path**: Where to save (default: `./generated-media/`)
• **Count**: How many images (default: 1)

**If user mentions "nano banana"** — they mean this built-in capability. Explain available options and proceed.

**Quality modes**:

• **Standard** (default): Cost-optimized chain (Gemini Flash → Gemini Pro → Pollinations → Imagen 4)
• **High**: Quality-optimized chain (Imagen 4 → Gemini Pro → Gemini Flash → Pollinations). Inform user: "Using high-quality mode — Imagen 4 costs ~$0.04/image."

**Resolution modes**:

• **Standard**: Default model output
• **2K**: Request 2048px detail — instruct via prompt suffix `", ultra detailed, 2048px quality"`
• **4K**: Request maximum quality — instruct via prompt suffix `", maximum quality 4K ultra detailed, sharp text rendering, 3840px"`

Step 2: Prepare Output Directory

mkdir -p ./generated-media

Step 3: Load API Key

# Source .env if it exists
if [ -f .env ]; then
  export $(grep -E '^GEMINI_API_KEY=' .env | xargs)
fi

# Check parent dirs (monorepo support)
if [ -z "$GEMINI_API_KEY" ] && [ -f ../.env ]; then
  export $(grep -E '^GEMINI_API_KEY=' ../.env | xargs)
fi

# Also load POLLINATIONS_API_KEY if available
if [ -f .env ]; then
  export $(grep -E '^POLLINATIONS_API_KEY=' .env | xargs 2>/dev/null) 2>/dev/null || true
fi

Step 4: Set Generation Parameters

TIMESTAMP=$(date +%s)
PROMPT="YOUR_PROMPT_HERE"  # The full prompt (with resolution suffix if 2K/4K)
ASPECT_RATIO="1:1"         # Set from user request (default: 1:1)
OUTFILE="generated-media/image-${TIMESTAMP}.png"
TMPFILE="/tmp/gemini-img-response-${TIMESTAMP}.json"
IN