# Bulk Diffusion 0.1a

**Massive and Continuous Picture Rendering using Stable Diffusion 1.5**

---

## Author
Hermann Knopp  
Contact: [[email protected]](mailto:[email protected])  

---

## Version
0.1a — 31.07.2023

---

## Overview
Bulk Diffusion is a Python application that generates images in bulk using **Stable Diffusion** and a wordlist-driven prompt system. It can automatically create diverse image prompts using English word lists, categories, and moods, optionally enhanced with **GPT-Neo** for more creative prompts.  

The software is designed for NVIDIA GPUs with at least 6GB VRAM. CPU-only rendering is **not supported yet**.

---

## Features

- Generates images continuously in bulk.
- Uses **Stable Diffusion 1.5** or other diffusion models (Dreamlike Diffusion, Inkpunk).
- Wordlist-based prompt generation:
  - `words.txt` — general descriptive words
  - `categories.txt` — picture categories
  - `moods.txt` — artistic moods
- Optional GPT-Neo prompt enhancement for more creative and stylistic outputs.
- Saves generated images and prompt details automatically in a specified folder.
- Interactive selection of model, output folder, and prompt engine.

---

## Installation

### Requirements
- Python 3.10.10 x64
- NVIDIA GPU with CUDA support (minimum 6GB VRAM)
- Virtual environment recommended (`venv`)

```bash
python -m venv venv
source venv/bin/activate  # Linux/macOS
venv\Scripts\activate     # Windows
pip install -r requirements.txt

---

Usage

1. Run the main Python script:

python bulk_diffusion.py

2. Follow interactive prompts:

   • Select the image output folder
   • Choose whether to use the GPT prompt engine
   • Choose the diffusion model (1–7)
   • Start batch rendering

3. Generated images and corresponding prompt text files will be saved automatically.

---

Supported Models

For GPTNeo Custom models, download the release from GitHub (pytorch_model.bin in model/ folder, ~500MB).

---

Output

• Images saved as .png
• Prompt text saved as .txt alongside each image
• Automatic naming: test_DDMMYYYY_HHMMSS.png and .txt
• Stores seed, category, mood, image size, guidance scale, and prompt

---

Hardware Requirements

• NVIDIA GPU with 6GB+ VRAM

• VRAM usage:

  • 4GB (FP32) / 2GB (FP16) for Stable Diffusion
  • 0.5GB for GPT-Neo model
  • 1–1.5GB RAM for PyTorch system files

• CPU-only rendering is not supported yet.

---

Wordlists

• words.txt — descriptive words for random prompt generation
• categories.txt — categories for image themes
• moods.txt — creative style or mood descriptors

All files can be edited and expanded by appending new words.

---

License

MIT License (or specify your preferred license)

---

Contact

GitHub Issues for bug reports or feature requests Email: [email protected]