ComfyUI-DJZ-ZeroEDIT

DJZ-ZeroEDIT

Procedural Semantic EDIT Prompt Generation for ComfyUI

A deterministic procedural system for generating AI image editing prompts using the ZeroBytes position-is-seed methodology. Perfect for batch editing workflows, A/B testing edit variations, and exploring the edit prompt space systematically.

Features

• Deterministic Generation: Same (seed, index, profile) → same edit prompt, always
• O(1) Performance: Instant generation via coordinate hashing, no iteration required
• JSON Profile System: Customizable vocabulary pools for different editing domains
• Edit-Specific Templates: Pre-structured for action, integration, and preservation clauses
• 4.4+ Billion Combinations: Default profile covers comprehensive editing scenarios

Installation

1. Clone or copy to your ComfyUI custom_nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/MushroomFleet/ComfyUI-ZeroEDIT-nodes

2. Install dependencies:

pip install xxhash

3. Restart ComfyUI

Nodes

DJZ Zero EDIT

Main generator node. Produces a single deterministic edit prompt.

Inputs:

• profile - JSON profile selection (from /profiles/edit/)
• seed - World seed (0 to 4,294,967,295)
• prompt_index - Position in infinite prompt space
• prefix (optional) - Text prepended to output
• suffix (optional) - Text appended to output

Output:

• edit_prompt - Complete edit instruction string

DJZ Zero EDIT Batch

Generates multiple sequential edit prompts for batch operations.

Inputs:

• profile - JSON profile selection
• seed - World seed
• start_index - Starting position
• batch_size - Number of prompts (1-64)
• prefix/suffix (optional)

Output:

• edit_prompts - List of edit instruction strings

DJZ Zero EDIT Profile Info

Displays statistics about the selected profile.

Output:

• info - Formatted string with pool sizes and combination count

Profile Structure

Profiles are JSON files in /profiles/edit/ containing:

{
  "name": "Profile Name",
  "description": "What this profile generates",
  "version": "1.0.0",
  "templates": [
    "Template with {pool_name} placeholders"
  ],
  "pools": {
    "pool_name": ["item1", "item2", "..."]
  }
}

Default Profile Pools

| Pool | Purpose | Count | |------|---------|-------| | edit_operation | Complete edit instructions (add/remove/change/transform) | 63 | | integration_detail | How changes should blend with scene | 12 | | blending_phrase | Seamlessness descriptions | 10 | | lighting_consistency | Light matching requirements | 10 | | shadow_detail | Shadow integration rules | 6 | | material_consistency | Surface/texture matching | 10 | | preservation_clause | What to keep unchanged | 14 | | style_element | Artistic style continuity | 10 |

Edit Operation Categories

The default profile covers these edit types:

Element Addition - Adding accessories, objects, environmental details

add a small knitted wizard hat on the subject's head...

Element Removal - Deleting unwanted objects, cleaning backgrounds

remove the person visible in the background, naturally reconstructing...

Attribute Modification - Changing colors, materials, time of day

change the dress to deep crimson red while maintaining the same cut...

Style Transfer - Converting to artistic styles

transform the photograph into the artistic style of Vincent van Gogh...

Multi-Image Composition - Combining elements from references

place the logo from the reference image centered on the chest...

Creating Custom Profiles

1. Create a new JSON file in /profiles/edit/
2. Define templates with {pool_name} placeholders
3. Populate pools with domain-specific vocabulary
4. Ensure all referenced pools exist

Example: Portrait Retouching Profile

{
  "name": "Portrait Retouch",
  "description": "Professional portrait editing operations",
  "version": "1.0.0",
  "templates": [
    "{retouch_action}. {skin_handling}. {preservation}."
  ],
  "pools": {
    "retouch_action": [
      "Smooth skin texture while preserving natural pores",
      "Brighten the eyes and add subtle catchlights",
      "Remove temporary blemishes from the face"
    ],
    "skin_handling": [
      "Maintain realistic skin texture and undertones",
      "Keep natural color variation in the skin"
    ],
    "preservation": [
      "Preserve the subject's unique features and character",
      "Keep the natural expression unchanged"
    ]
  }
}

ZeroBytes Methodology

This node implements position-is-seed procedural generation:

• Coordinate System: (seed, prompt_index, component_index)
• Hash Function: xxhash32 for speed and determinism
• No Iteration: Any prompt accessible in O(1) time
• Cross-Platform: Same results on any system

# Conceptual model
hash(seed, prompt_idx=1000, component=0) → template selection
hash(seed, prompt_idx=1000, component=1) → edit_operation selection
hash(seed, prompt_idx=1000, component=2) → integration_detail selection
# ...

Workflow Examples

Sequential Edit Exploration

Seed: 42
Index 0 → "add a small knitted wizard hat..."
Index 1 → "turn this rough sketch into..."
Index 2 → "add a colorful silk scarf..."

Parallel Seed Exploration

Index: 100
Seed 0 → variation A
Seed 1 → variation B
Seed 2 → variation C

Batch Processing

Use the Batch node to generate 64 variations at once for automated testing.

Combination Mathematics

Total unique prompts = templates × pool₁ × pool₂ × ... × poolₙ

Default profile:

7 × 63 × 12 × 10 × 10 × 6 × 10 × 14 × 10 = 4,445,280,000

That's over 4.4 billion unique edit prompts from a single profile.

License

MIT License - See LICENSE file

Credits

Based on DJZ-ZeroPrompt-V2 architecture. Edit prompt patterns derived from Nano Banana Pro 2026 documentation and FLUX.1 Kontext community guidance.

📚 Citation

Academic Citation

If you use this codebase in your research or project, please cite:

@software{comfyui_zeroedit_nodes,
  title = {ComfyUI-ZeroEDIT-nodes: Procedural Semantic EDIT Prompt Generation for ComfyUI},
  author = {Drift Johnson},
  year = {2025},
  url = {https://github.com/MushroomFleet/ComfyUI-ZeroEDIT-nodes},
  version = {1.0.0}
}

Donate: