Extensions/ComfyUI-DJZ-ZeroPrompt
ComfyUI Extension

ComfyUI-DJZ-ZeroPrompt

Zero files. Zero storage. Infinite prompts

By MushroomFleet·Created 7 months ago·Updated 4 months ago· 2
MushroomFleet/ComfyUI-DJZ-ZeroPrompt
Nodes—
On cloudLocal install
Stars2
Updated4 months ago
Readme

ComfyUI-DJZ-ZeroPrompt

Zero files. Zero storage. Infinite prompts.

Part of the DJZ-Zerobytes 0(1) procedural collection

  • https://github.com/MushroomFleet/ComfyUI-DJZ-ZeroPrompt
  • https://github.com/MushroomFleet/ComfyUI-ZeroENH
  • https://github.com/MushroomFleet/ComfyUI-ZeroEDIT-nodes

A ComfyUI custom node pack that generates deterministic prompts using position-is-seed procedural generation. Instead of reading from user-maintained text files, these nodes compute prompts directly from mathematical coordinates—the same seed and index will always produce the same prompt, everywhere, every time.

ZeroPrompt Banner ComfyUI License

🚀 Features

  • Infinite Prompts: 188+ trillion unique prompt combinations from the default profile
  • Perfect Determinism: seed=42, index=1000 produces identical output across all machines
  • O(1) Generation: Direct hash computation—no file I/O, no iteration
  • Zero Maintenance: No text files to curate, organize, or update
  • Reproducible Results: Share seed+index coordinates to recreate exact prompts
  • Custom Profiles (V2): JSON-based vocabulary profiles for specialized generation

📦 Installation

Method 1: ComfyUI Manager (Recommended)

  1. Open ComfyUI Manager
  2. Search for "DJZ-ZeroPrompt"
  3. Click Install

Method 2: Manual Installation

cd ComfyUI/custom_nodes
git clone https://github.com/MushroomFleet/ComfyUI-DJZ-ZeroPrompt
cd ComfyUI-DJZ-ZeroPrompt
pip install -r requirements.txt

Method 3: Direct Download

  1. Download the repository as ZIP
  2. Extract to ComfyUI/custom_nodes/ComfyUI-DJZ-ZeroPrompt
  3. Install dependencies: pip install xxhash

🎯 Nodes

DJZ Zero Prompt V1

The original node with built-in vocabulary. Simple and fast.

| Input | Type | Description | |-------|------|-------------| | seed | INT | World seed (0 to 4,294,967,295) | | prompt_index | INT | Position in prompt space | | prefix | STRING | Optional text to prepend | | suffix | STRING | Optional text to append |

DJZ Zero Prompt V2

Enhanced node with JSON profile support for customizable vocabulary.

| Input | Type | Description | |-------|------|-------------| | profile | DROPDOWN | Select vocabulary profile | | seed | INT | World seed (0 to 4,294,967,295) | | prompt_index | INT | Position in prompt space | | prefix | STRING | Optional text to prepend | | suffix | STRING | Optional text to append |

DJZ Zero Prompt Profile Info

Utility node to display profile statistics (pool sizes, total combinations).

📁 Profile System (V2)

V2 introduces JSON profiles for customizable prompt generation. Profiles are stored in the /profiles folder.

Included Profiles

| Profile | Description | Combinations | |---------|-------------|--------------| | default.json | Full vocabulary, all categories | 188+ trillion | | cyberpunk.json | Neon-soaked dystopian sci-fi | 31+ billion | | fantasy.json | Epic high fantasy and medieval | 4+ billion |

Creating Custom Profiles

  1. Copy an existing profile from /profiles
  2. Rename it (e.g., my_style.json)
  3. Edit the vocabulary pools and templates
  4. Restart ComfyUI—your profile appears in the dropdown

Profile JSON Structure

{
  "name": "My Custom Profile",
  "description": "Description of the style",
  "version": "1.0.0",
  
  "templates": [
    "{subject} {action} {environment}, {style}, {lighting}",
    "{style} {subject} in {environment}, {mood} atmosphere"
  ],
  
  "pools": {
    "subject": ["a warrior", "a wizard", "a robot"],
    "action": ["standing in", "exploring", "fighting in"],
    "environment": ["a forest", "a city", "a cave"],
    "style": ["photorealistic", "anime style", "oil painting"],
    "lighting": ["golden hour", "neon lighting", "moonlight"],
    "mood": ["epic", "mysterious", "serene"]
  }
}

Template Variables

Templates use {pool_name} syntax. Available variables depend on the pools defined:

  • {subject} - Character or entity
  • {action} - What they're doing
  • {environment} - Where they are
  • {style} - Art style
  • {lighting} - Light conditions
  • {camera} - Shot type
  • {details} - Quality modifiers
  • {mood} - Atmosphere

You can add custom pools—just define them in pools and reference them in templates.

💡 How It Works

The ZeroBytes Methodology

Traditional prompt systems store prompts in files and select randomly. This approach has limitations:

  • Finite prompts (limited by file size)
  • Maintenance burden (curating text files)
  • Non-deterministic selection (different results on different runs)

ZeroPrompt uses position-is-seed procedural generation:

(seed, prompt_index, [profile]) → xxhash32 → deterministic component selection → formatted prompt

A prompt is not retrieved—it's computed from coordinates in semantic space.

Architecture

Prompt Space Hierarchy:
━━━━━━━━━━━━━━━━━━━━━━━
seed (world seed)
  └─ prompt_index (position)
       ├─ template[hash(seed, idx, 0)]
       ├─ subject[hash(seed, idx, 1)]
       ├─ action[hash(seed, idx, 2)]
       ├─ environment[hash(seed, idx, 3)]
       ├─ style[hash(seed, idx, 4)]
       ├─ lighting[hash(seed, idx, 5)]
       ├─ camera[hash(seed, idx, 6)]
       ├─ details[hash(seed, idx, 7)]
       └─ mood[hash(seed, idx, 8)]

Vocabulary Pools (Default Profile)

| Pool | Entries | Examples | |------|---------|----------| | Subjects | 88 | "a knight", "a cyborg", "a dragon", "a witch" | | Actions | 51 | "standing in", "exploring", "battling through" | | Environments | 56 | "a dark forest", "a cyberpunk city", "the void" | | Styles | 58 | "photorealistic", "anime style", "oil painting" | | Lighting | 35 | "golden hour", "neon lighting", "volumetric fog" | | Camera | 31 | "close-up portrait", "wide shot", "dutch angle" | | Details | 31 | "highly detailed", "8k", "masterpiece" | | Mood | 48 | "serene", "ominous", "mysterious", "epic" | | Templates | 8 | Various composition structures |

Total combinations: 188,274,509,660,160 (188+ trillion)

📊 Example Outputs

Default Profile (seed=42)

[0] a dwarven smith, the astral plane, hyperrealistic, blue hour lighting, 
    intense atmosphere, intricate details

[1] a wolf in inside a computer mainframe, timeless vector art, sunset backlight, 
    point of view shot, pristine quality

Cyberpunk Profile (seed=42)

[0] cinematic bokeh background, a data thief making deals in a augmentation clinic, 
    cinematic, pulsing neon, mysterious

[3] oppressive scene of a chrome-armed warrior running through a skybridge, 
    dystopian realism, volumetric fog, realistic textures

Fantasy Profile (seed=42)

[0] a noble prince standing guard in a mystical lake, epic fantasy art, 
    moonlight, medium shot, masterfully crafted, triumphant atmosphere

[2] lord of the rings style a fearsome werewolf in a garden of statues, 
    aurora borealis, mysterious mood, 8k resolution

🔄 Determinism Guarantee

The same (seed, prompt_index, profile) tuple will always produce the same prompt:

# These will always be identical
prompt_a = generate(seed=42, idx=1000, profile="default.json")  # Run 1
prompt_b = generate(seed=42, idx=1000, profile="default.json")  # Run 2
prompt_c = generate(seed=42, idx=1000, profile="default.json")  # Different machine

assert prompt_a == prompt_b == prompt_c  # Always True

This enables:

  • Reproducible workflows: Save seed+index+profile instead of prompt text
  • Collaboration: Share coordinates, not strings
  • Version control: Track prompt changes via seed/index history

🛠️ Technical Details

Dependencies

  • xxhash - Fast, deterministic hashing (xxhash32)
  • Python 3.8+

Why xxhash32?

| Property | xxhash32 | Python random | hash() | |----------|----------|-----------------|----------| | Deterministic | ✓ | ✓ (with seed) | ✗ (randomized per session) | | Cross-platform | ✓ | Edge cases | ✗ | | O(1) direct access | ✓ | ✗ (iterates) | ✓ | | Designed for this | ✓ | ✗ | ✗ |

Performance

  • Generation time: <1ms per prompt
  • Memory: ~50KB (vocabulary pools in code/JSON)
  • No I/O: Pure computation (V1) or single JSON read (V2, cached)

🎨 Workflow Tips

Exploring Prompt Space

  • Fixed seed, varying index: Explore one "universe" of prompts
  • Varying seed, fixed index: See how the same "position" differs across universes
  • Random seed + random index: Maximum variety
  • Different profiles: Same seed+index, different vocabulary = different genre

Finding Good Prompts

  1. Start with seed=0, index=0
  2. Increment index to browse prompts
  3. When you find a style you like, note the seed
  4. Explore nearby indices for variations
  5. Try different profiles for genre variations

Batch Generation

Connect to a primitive that increments prompt_index to generate batches:

prompt_index = batch_number * batch_size + item_index

📁 Repository Structure

ComfyUI-DJZ-ZeroPrompt/
├── __init__.py              # ComfyUI node registration
├── DJZ_ZeroPrompt_V1.py     # V1 node (built-in vocabulary)
├── DJZ_ZeroPrompt_V2.py     # V2 node (JSON profiles)
├── profiles/                # Vocabulary profiles
│   ├── default.json         # Full vocabulary
│   ├── cyberpunk.json       # Cyberpunk/sci-fi focused
│   └── fantasy.json         # High fantasy focused
├── requirements.txt         # Python dependencies
├── README.md                # This file
└── LICENSE                  # MIT License

🤝 Contributing

Contributions welcome! Areas of interest:

  • Additional vocabulary entries
  • New profile themes (horror, romance, nature, etc.)
  • Template structure improvements
  • Performance optimizations
  • Documentation improvements

📜 License

MIT License - See LICENSE for details.

📚 Citation

Academic Citation

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

@software{djz_zeroprompt,
  title = {ComfyUI-DJZ-ZeroPrompt: Procedural Semantic Prompt Generation using Position-is-Seed Methodology},
  author = {Drift Johnson},
  year = {2025},
  url = {https://github.com/MushroomFleet/ComfyUI-DJZ-ZeroPrompt},
  version = {1.0.0}
}

Donate

Ko-Fi

Support This Project

If you found this useful, please star the repo — it helps others discover it!

Star on GitHub