EricRollei-Local_LLM_Prompt_Enhancer/docs/QUICK_REF_V17.md

Text-to-Image Node v1.7 - Quick Reference

What's New

🔧 Fixed

• Reference Images: Now properly analyzed and incorporated into prompts
  • Analyzes: dimensions, brightness, color tones
  • LLM receives detailed characteristics to match

✨ New Controls

Prompt Length

Control the output size:

• very_short - 20-40 tokens (minimal)
• short - 40-80 tokens (SDXL)
• medium - 80-150 tokens (Flux)
• long - 150-250 tokens (complex)
• very_long - 250-400 tokens (maximum detail)

Genre/Style

Infuse prompts with specific mood:

• Creative: surreal, cinematic, dramatic, action, artistic
• Horror: horror, noir
• Tech: scifi, cyberpunk
• Mood: humorous, romantic, indie
• Era: vintage, modern
• Content: pg, x-rated
• Aesthetic: minimalist, maximalist, documentary

Subject Framing

Control the shot composition:

• extreme close-up - Facial details
• close-up - Face and shoulders
• medium shot - Waist up
• wide shot - Full body with environment
• full body - Complete subject
• cowboy shot - Mid-thigh up
• Plus: bust shot, head and shoulders, three-quarter

Subject Pose

Control character positioning:

• Static: standing, sitting, lying down, kneeling
• Dynamic: action pose, walking, running, jumping, dancing
• Portrait: portrait pose, contrapposto, asymmetric
• Energy: relaxed, tense, dynamic, static

---

Quick Examples

Example 1: Cinematic Portrait

Prompt: "woman in evening dress"
Genre: cinematic
Length: medium
Framing: close-up
Pose: contrapposto

→ Film-quality portrait, dramatic lighting, classic pose

Example 2: Horror Scene

Prompt: "abandoned mansion"
Genre: horror
Length: long
Framing: wide shot

→ Detailed atmospheric horror with ominous mood

Example 3: Action Shot

Prompt: "superhero"
Genre: action
Length: short
Framing: wide shot
Pose: dynamic

→ Fast-paced, energetic, movement-focused

Example 4: Cyberpunk Character

Prompt: "{male|female} hacker"
Genre: cyberpunk
Length: medium
Framing: medium shot
Pose: asymmetric

→ Neon, tech, gritty urban dystopia aesthetic

---

Pro Tips

Reference Images

• Use clear, well-lit references for best results
• Color/lighting characteristics transfer well
• Works with any ComfyUI-compatible image format

Combining Features

• Alternation {a|b|c} + Emphasis (keyword:1.5) + Genre + Framing = powerful control
• All features work together seamlessly
• Random options provide variety across generations

Platform Specific

• SDXL: Use short length, works with all genres
• Flux: medium length, natural language excels
• Pony: Any length converts to tags, genres adapt
• Chroma: long or very_long for complex scenes

Length vs Detail

• Shorter = focused, concise, faster to generate
• Longer = complex scenes, multiple subjects, rich atmosphere
• Match length to scene complexity

---

All New Inputs

Input	Options	Default
genre_style	22 options	auto
prompt_length	6 options	auto
subject_framing	14 options	auto
subject_pose	17 options	auto

All support:

• auto - Intelligent defaults
• random - Variety across generations
• none - Disable the control
• Specific values for precise control

---

Testing Checklist

• Reference image incorporated (check output mentions image characteristics)
• Prompt length matches target (count tokens ~words × 1.3)
• Genre style reflected (vocabulary/tone matches)
• Framing mentioned (shot type present)
• Pose described (body position clear)
• All features work together
• Platform formatting correct
• Emphasis syntax preserved (keyword:1.5)
• Alternation processed {a|b|c}
• No settings leakage in output

---

Restart Required

After update, restart ComfyUI completely to load new controls!

---

Files Updated

• text_to_image_node.py - Core functionality
• UPDATE_V17_ENHANCED_CONTROLS.md - Full documentation
• This file - Quick reference

Version: 1.7
Status: ✅ Ready to use
Compatibility: ✅ All existing workflows work