Ai Image Generation

Generate images using pre-configured prompts and default settings.

Use when:

• Creating standard book illustrations
• Need quick visualization of a scene
• Working with established style guide

Commands:

# Generate cover
npm run generate:cover

# Generate chapter illustrations
npm run generate:illustrations

# Single illustration
node scripts/generate-single.js mid

MCP Usage:

@replicate run black-forest-labs/flux-schnell with prompt "boy with magnifying glass"

Level 2: Advanced Generation

Create variations, compare models, test character consistency, and customize prompts.

Use when:

• Need multiple options to choose from
• Comparing quality across models
• Testing character consistency for multi-scene projects
• Fine-tuning prompt for specific results

Commands:

# Generate 3 variations
node scripts/generate-variations.js --type=cover --model=flux-pro --count=3

# Test character consistency (6 scenes × 2 styles)
node scripts/generate-consistency-test.js

# Compare models
node scripts/generate-cover.js --model=flux-schnell
node scripts/generate-cover.js --model=flux-pro
node scripts/generate-cover.js --model=ideogram

# Custom prompt
node scripts/generate-single.js custom --prompt="your prompt here"

MCP Usage:

@replicate search "flux image generation"
@replicate run black-forest-labs/flux-1.1-pro with prompt "..."

Level 3: Expert Workflows

Fine-tune custom models, batch process, and integrate into publishing pipeline.

Use when:

• Building consistent visual brand
• Processing entire book at once
• Creating custom fine-tuned models

Capabilities:

• Fine-tune FLUX LoRA on your art style
• Batch process all chapters
• Upscale with super-resolution models
• Generate video from illustrations
• Custom deployment for production

Resources: See resources/ for:

• Fine-tuning guides
• Batch processing templates
• Model comparison matrices
• Cost optimization strategies

Models Reference

*Validated finding: FLUX Schnell can produce production-quality work for simple compositions (portraits, clear expressions)

Style Guidelines

CRITICAL: Discover user's aesthetic preference first:

• Don't assume B&W line art for middle-grade books
• Generate 2-3 style samples (realistic vs line art)
• Let user explicitly choose preferred aesthetic
• Adjust model strategy accordingly

For realistic/photorealistic style:

photorealistic 3D render, atmospheric cinematic lighting,
professional middle-grade book illustration,
vibrant color palette, dramatic composition

For B&W line art style:

black and white line art, middle-grade book illustration, 
clean lines, no shading, suitable for ages 8-12

For covers (add to style above):

professional book cover design, vertical orientation (2:3), 
dramatic composition, sense of mystery and intelligence

Common Patterns

Discover style preference (do this first!):

# Generate realistic sample
node scripts/generate-cover.js --model=flux-pro
# Generate line art sample  
node scripts/generate-single.js opening --model=flux-schnell
# User chooses → document preference

Test character consistency (for multi-scene books):

node scripts/generate-consistency-test.js
# Review outputs → Rate consistency 1-5
# 4-5 stars = viable for book illustration

Generate cover variations:

node scripts/generate-variations.js --type=cover --count=5

Generate all chapter 1 scenes:

node scripts/generate-illustrations.js 1

Quick test (may be production-worthy!):

node scripts/generate-single.js opening --model=flux-schnell

Atmospheric/complex scenes:

node scripts/generate-variations.js --model=flux-pro --count=3

Troubleshooting

Rate limiting (429 error):

• Script auto-waits and retries
• With <$5 credit: 6 requests/minute
• Add more credit for higher limits

No output received:

• FLUX Pro may return different format
• Script handles this automatically
• Check assets/variations/ for output

Insufficient credit (402 error):

• Visit https://replicate.com/account/billing
• Add $10+ credit (gets ~250 FLUX Pro images)
• Wait 2-3 minutes for activation

Cost Optimization

For testing: Use FLUX Schnell ($0.003) For review: Generate 3-5 variations, pick best For production: Use FLUX Pro only on final selected prompts For covers with text: Try Ideogram if text is critical

Estimated costs:

• Full book (30 chapters × 3 illustrations): $0.27 (Schnell) or $3.60 (Pro)
• Cover variations (5 options): $0.20 (Pro)
• Character portraits (10 variations): $0.40 (Pro)

Related Skills

• markdown-mermaid: Diagrams and flowcharts
• svg-graphics: Vector illustrations (free, full control)
• brand-asset-management: Logo and visual identity

Next Steps

1. Generate first image: npm run generate:cover
2. Review output in assets/
3. Try variations to compare: node scripts/generate-variations.js --count=3
4. Choose best and iterate on prompt
5. Scale to full book production

Documentation: docs/IMAGE-GENERATION-SETUP.md