Stats
Actions
Tags
From creative
Execute AI image generation with optimal quality. Use when you need to generate images via Replicate API. Triggers on: generate image, create visual, product shot. Outputs generated images for feedback and iteration.
How this skill is triggered — by the user, by Claude, or both
Slash command
/creative:ai-image-generationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This is the core execution skill for AI image generation. It handles the actual generation work — model selection, prompt construction, API calls, iteration, and quality verification.
This is the core execution skill for AI image generation. It handles the actual generation work — model selection, prompt construction, API calls, iteration, and quality verification.
Use this skill when: You need to generate any image and want optimal quality. Route here from: ai-creative-workflow, ai-creative-strategist, or direct requests.
The problem: Most people waste time and money on AI image generation because they:
The solution: A systematic approach that:
Why: Fastest high-quality generation available. ~20 seconds vs ~3 minutes for comparable quality.
| Model | Glif ID | Speed | Quality |
|---|---|---|---|
| 🍌 Nano Banana Pro Text 2 Image | cmi7ne4p40000kz04yup2nxgh | ~20sec | Excellent |
Use the run_glif MCP tool:
Glif: 🍌 Nano Banana Pro Text 2 Image
ID: cmi7ne4p40000kz04yup2nxgh
Inputs: ["Your prompt here"]
Do NOT use: imagen-3, DALL-E, Flux, or direct Replicate API calls unless specifically requested.
IF user specifically requests a different model:
use requested model
ELIF need text rendering (logos, quotes, announcements):
use Recraft v3 or Ideogram v3 (route to ai-typography)
ELIF need image-to-video:
generate image first, then route to ai-video-generation
ELIF need people/lifestyle imagery:
still use Nano Banana Pro (handles people well)
ELSE:
Nano Banana Pro via Glif // DEFAULT
[SUBJECT] + [SETTING/CONTEXT] + [STYLE] + [LIGHTING] + [QUALITY BOOSTERS]
Each element serves a purpose:
Bad: "A coffee cup" Good: "Artisan ceramic coffee cup with latte art, steam rising"
Bad: "A laptop" Good: "Silver MacBook Pro on oak desk, screen showing code editor"
Bad: "A person working" Good: "Woman in her 30s, focused expression, typing on laptop in bright modern office"
The more specific your subject, the less the model has to guess.
Studio/Clean:
on pure white background
against dark charcoal gradient
floating in negative space
on seamless backdrop
Environmental:
in modern minimalist office
on rustic wooden table
in cozy coffee shop
against urban cityscape
Abstract:
with geometric patterns
surrounded by abstract shapes
with dynamic particle effects
in surreal dreamscape
Photography Styles:
professional product photography
lifestyle photography
editorial fashion photography
documentary photography style
portrait photography
macro photography
aerial/drone photography
Illustration Styles:
digital illustration
watercolor illustration
vector art style
hand-drawn sketch style
3D rendered
isometric illustration
flat design
Artistic Styles:
cinematic
surrealist
minimalist
maximalist
retro/vintage
futuristic
organic/natural
Lighting is the most underused prompt element. It dramatically affects mood and quality.
Studio Lighting:
professional studio lighting
three-point lighting setup
single dramatic spotlight
soft diffused lighting
rim lighting
Natural Lighting:
soft natural daylight
golden hour light
overcast diffused light
harsh midday sun
blue hour
Dramatic Lighting:
high contrast dramatic lighting
moody low-key lighting
backlit silhouette
chiaroscuro lighting
neon accent lighting
Specific Directions:
key light from upper left
side lighting from right
overhead soft box
underlighting (dramatic)
Add these to elevate output quality:
Technical Quality:
ultra-detailed
high resolution
sharp focus
professional quality
8k resolution
Professional Context:
commercial photography
award-winning
magazine quality
advertising campaign
editorial quality
Specific Techniques:
shallow depth of field
bokeh background
lens flare
motion blur
long exposure
Premium wireless headphones floating against dark charcoal gradient,
professional product photography, single dramatic spotlight from upper left,
subtle rim lighting on edges, metallic surface reflections,
ultra-detailed, commercial advertising quality, negative space on right for text
Young professional woman enjoying morning coffee in bright modern apartment,
natural window light streaming in, cozy weekend aesthetic,
lifestyle photography, warm tones, genuine relaxed expression,
shallow depth of field, editorial quality
Gourmet burger with melted cheese and fresh vegetables on rustic wooden board,
professional food photography, soft diffused overhead lighting,
steam rising, appetizing presentation, shallow depth of field,
restaurant menu quality, warm inviting tones
Modern dashboard interface on large monitor in minimalist home office,
clean UI design visible on screen, soft ambient lighting,
tech product photography, contemporary workspace aesthetic,
shallow depth of field on screen, professional quality
Abstract representation of data flow and connectivity,
glowing particles forming network patterns against deep blue gradient,
futuristic technology aesthetic, dynamic motion blur,
cinematic quality, vibrant cyan and purple accent colors
| Use Case | Ratio | Pixels | Notes |
|---|---|---|---|
| Square (Instagram, LinkedIn) | 1:1 | 1024×1024 | Most versatile |
| Landscape (Hero banners, YouTube) | 16:9 | 1280×720 | Web standard |
| Portrait tall (Instagram feed) | 4:5 | 1024×1280 | Better feed presence |
| Stories/TikTok | 9:16 | 1080×1920 | Vertical mobile |
| 2:3 | 1000×1500 | Optimal for pins | |
| Wide cinematic | 21:9 | 2560×1080 | Dramatic headers |
Default: 1:1 unless platform-specific requirement is known.
Before generating, know:
□ What is the subject? (be specific)
□ What style/aesthetic?
□ What platform/use case?
□ What aspect ratio?
□ Any must-have elements?
□ Any must-avoid elements?
Use the formula:
[SUBJECT] + [SETTING] + [STYLE] + [LIGHTING] + [QUALITY BOOSTERS]
Glif: 🍌 Nano Banana Pro Text 2 Image
ID: cmi7ne4p40000kz04yup2nxgh
Inputs: ["[constructed prompt]"]
## Image Generated
**Prompt:** [prompt used]
**Aspect Ratio:** [ratio]
**Image URL:** [URL]
**Quick Assessment:**
- Subject accuracy: ✓/✗
- Style match: ✓/✗
- Composition: ✓/✗
- Quality: ✓/✗
**Feedback?**
- What works?
- What should change?
- Ready to approve or iterate?
Common iteration patterns:
"Too dark" → Add "bright" "well-lit" "high key lighting"
"Too busy" → Add "minimal" "clean" "simple composition"
"Wrong style" → Be more explicit about style reference
"Missing element" → Add specific element to prompt
"Wrong mood" → Adjust lighting and color descriptors
Problem: Almost right but something's off Strategy: Make targeted prompt adjustments
Original: "Coffee cup on table"
Iteration 1: "Coffee cup on table, warmer tones" (if too cool)
Iteration 2: "Coffee cup on table, warmer tones, more steam" (if missing element)
Problem: Output doesn't match intent at all Strategy: Rewrite prompt from scratch with more specificity
Don't iterate on a bad foundation. Start fresh with better prompt construction.
Problem: Image looks AI-generated or low quality Strategy: Add quality boosters and professional context
Add: "ultra-detailed, professional photography, commercial quality"
Add: "8k resolution, sharp focus, magazine quality"
Problem: Elements in wrong places, bad framing Strategy: Be explicit about composition
Add: "subject centered in frame"
Add: "negative space on left for text"
Add: "rule of thirds composition"
Add: "close-up shot" or "wide establishing shot"
Before delivering, verify:
| Issue | Cause | Solution |
|---|---|---|
| Generic/stock photo feel | Vague prompt | Add specific details, unique elements |
| Wrong style | Style not specified | Be explicit: "product photography style" not just "photo" |
| Poor lighting | No lighting direction | Add specific lighting: "dramatic spotlight from upper left" |
| Cluttered composition | Too many elements | Add "minimal" "clean" "simple composition" |
| AI artifacts | Model limitations | Regenerate, or note area to avoid |
| Wrong aspect ratio | Default used | Specify ratio in request |
| Doesn't match brand | No style reference | Add brand color/style keywords |
| Text rendering issues | Wrong model | Route to ai-typography for text-heavy graphics |
## Image Generated
**Subject:** [what was generated]
**Style:** [style used]
**Aspect Ratio:** [ratio]
**Image URL:** [URL]
**Prompt Used:**
> [full prompt for reference]
**Quality Check:**
- ✓ Subject accurate
- ✓ Style matches
- ✓ Composition works
- ✓ Quality sufficient
**Options:**
- [ ] Approve and use
- [ ] Iterate (specify changes)
- [ ] Generate variant
- [ ] Try different approach
## Image Generated (Iteration 2)
**Changes from v1:** [what changed]
**Image URL:** [URL]
**Comparison:**
- v1: [issue]
- v2: [improvement]
**Feedback needed?**
Receive:
subject: "[what to generate]"
style: "[aesthetic direction]"
platform: "[intended use]"
aspect_ratio: "[ratio]"
constraints: "[any limitations]"
Return:
status: "complete" | "needs_iteration" | "needs_different_approach"
deliverables:
- url: "[image URL]"
aspect_ratio: "[ratio]"
prompt_used: "[prompt]"
feedback_needed: "[any questions]"
Route to ai-video-generation:
image_url: "[approved image URL]"
aspect_ratio: "[ratio]"
animation_intent: "[what motion is needed]"
IMAGE GENERATION PIPELINE
┌─────────────────────────────────────────┐
│ Request arrives (direct or routed) │
│ → Clarify requirements │
│ → Construct prompt │
└─────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────┐
│ ai-image-generation (THIS SKILL) │
│ → Generate with Nano Banana Pro/Glif │
│ → Present for feedback │
│ → Iterate as needed │
│ → Quality check │
└─────────────────────────────────────────┘
│
┌───────────┴───────────┐
▼ ▼
┌──────────────────┐ ┌──────────────────┐
│ Static Delivery │ │ Route to Video │
│ → Final image │ │ → ai-video- │
│ delivered │ │ generation │
└──────────────────┘ └──────────────────┘
QUICK DRAFT (concept testing):
→ Nano Banana Pro via Glif (~20 sec)
→ Good enough to evaluate concept
→ Iterate quickly
PRODUCTION QUALITY (final delivery):
→ Nano Banana Pro via Glif (~20 sec)
→ Already production quality
→ Most cases don't need more
MAXIMUM QUALITY (premium needs):
→ Direct Replicate API call to specialized model
→ Only if Glif output insufficient
→ Rare — Nano Banana Pro handles most needs
Default to Glif. It's fast AND high quality. Only go elsewhere if specific need arises.
80% of good results come from:
Get these four right, and you'll get good results.
npx claudepluginhub 0xobat/claude-skills --plugin creativeGenerates AI images from text prompts, edits images, and composes from multiple references using Gemini models. Supports t2i, i2i, product mockups, and stickers.
Generates optimized prompts for Gemini 2.5 Flash Image (Nano Banana) using best practices for photorealistic shots, art styles, and multi-turn editing workflows.
Enhances image generation prompts with Subject-Context-Style structure, lighting physics, camera terminology, and character consistency patterns. Useful for creating detailed, physically coherent image prompts.