AI Artist - Prompt Engineering + Image Generation

Comprehensive prompt engineering with integrated image generation. Contains 6,000+ curated prompts, dual-option generation workflow, and searchable database with BM25 ranking.

Dual-Option Workflow

When generating images, always offer both options:

Option	Model	Best For	Speed
Standard	`gemini-2.5-flash-image`	Search-based prompts, known formats	Fast
Creative	`gemini-3-pro-image-preview`	Unique art direction, complex scenes	Best quality

Quick Start

Option 1: Standard (Search + Flash)

bash

# 1. Search for similar prompts
python3 .claude/skills/ai-artist/scripts/search.py "<keywords>" --domain examples

# 2. Generate with Flash model
.claude/skills/.venv/bin/python3 .claude/skills/ai-multimodal/scripts/gemini_batch_process.py \
  --task generate --model gemini-2.5-flash-image \
  --prompt "<adapted_prompt>" --aspect-ratio <ratio> --size 2K --output <path>

Option 2: Creative (Art Direction + Pro)

bash

# Generate with Pro model (4K, complex prompts)
.claude/skills/.venv/bin/python3 .claude/skills/ai-multimodal/scripts/gemini_batch_process.py \
  --task generate --model gemini-3-pro-image-preview \
  --prompt "<creative_prompt>" --aspect-ratio <ratio> --size 4K --output <path>

Arguments

Arg	Values	Default
`--model`	`gemini-2.5-flash-image` (Standard), `gemini-3-pro-image-preview` (Creative)	Flash
`--aspect-ratio`	`1:1`, `16:9`, `9:16`, `3:4`, `4:3`, `2:3`, `3:2`	`1:1`
`--size`	`1K`, `2K`, `4K` (4K only Pro)	`2K`
`--num-images`	`1-4`	`1`
`--output`	file path	auto

Search Domains

bash

python3 .claude/skills/ai-artist/scripts/search.py "<query>" --domain <domain>

Domain	Use For
`examples`	6000+ curated prompts (quote cards, bento grids, thumbnails)
`style`	Visual aesthetics (cyberpunk, minimalist, cinematic)
`subject`	Subject tips (portrait, product, landscape)
`platform`	Platform syntax (midjourney, dall-e, sd)

Creative Art Directions

For Creative option, apply unique styles:

Style	Key Elements
Cyberpunk Neon	Holographic, neon glow, rain, cityscape, Japanese text
Vaporwave	Greek statues, sunset grid, Windows 95, chrome, dolphins
Isometric 3D	Tilt-shift, miniature diorama, cute buildings, floating island
Bento Grid	Apple liquid glass, transparent cards, gradient glow
Editorial	Clean white, bold typography, magazine layout

Example Workflow

User: "Create marketing infographic"

Standard Option:

bash

python3 .claude/skills/ai-artist/scripts/search.py "bento grid marketing" --domain examples
# → Found: "Premium liquid glass Bento grid product infographic"
# → Adapt prompt, generate with Flash

Creative Option:

bash

# Cyberpunk art direction with Pro model
.claude/skills/.venv/bin/python3 .claude/skills/ai-multimodal/scripts/gemini_batch_process.py \
  --task generate --model gemini-3-pro-image-preview \
  --prompt "Create a cyberpunk neon holographic infographic titled 'MARKETING 2026'. Style: Blade Runner aesthetic, dark cityscape, rain, neon data cards floating around holographic AI head. Hyper-detailed, cinematic." \
  --aspect-ratio 1:1 --size 4K --output infographic-cyberpunk.png

References

Topic	File
Image generation workflow	`references/image-generation-workflow.md`
Nano Banana models	`references/nano-banana.md`
Prompt examples (6000+)	`references/nano-banana-pro-examples.md`
LLM patterns	`references/llm-prompting.md`
Image syntax	`references/image-prompting.md`
Domain patterns	`references/domain-patterns.md`

Anti-Patterns

Issue	Solution
Only offering one option	Always present Standard AND Creative
Using Flash for complex art	Use Pro for unique art direction
Vague creative prompts	Add detailed style, layout, elements
Missing output path	Always specify `--output`