I needed 10K+ unique images for my ML model - so I built PromptAnvil
I was training my own ML model and needed a large, diverse image dataset. Scraping from the internet? Too messy, copyright issues, inconsistent quality.
GPT? Not great at this scale - repetitive outputs, no structure, hard to categorize.
So I decided to generate my own images using AI. But here's the problem: I needed variety and categorization. Thousands of unique prompts, organized by subject, style, lighting, camera angle... manually writing them wasn't an option.
The Solution
That's why I built PromptAnvil. It lets me set up a "recipe" for my prompts - which subjects, styles, lighting setups I want, how often each should appear, and rules like "if it's an underwater scene, use caustic lighting." Then I just hit generate and get hundreds of unique, organized prompts ready for image generation.
Features:
Weighted randomization for controlled variety
Conditional logic rules (IF/THEN)
Tag linking - keep context coherent by grouping related entries across categories
Smart templates with
{token}systemExport to JSON/TXT/CSV for automation pipelines
Cloud sync & public config sharing
AI-powered helpers (in active development)
Works with Midjourney, Stable Diffusion, DALL-E, or any image AI.
Try it: https://promptanvil.com
Built this for myself, now sharing it with the community. Feedback welcome! 🙏
