Thank you for contributing. This library is for everyone, and it grows because real people send in the prompts that worked for them.
- Real, usable prompts only. No placeholders like
[subject here]. - Tool-agnostic wording. Use the language of photography, cinematography, or fine art. Do not name specific diffusion models.
- One prompt per entry. If you have a family of related prompts, submit them as separate entries.
- 1–2 sentence "Produces:" description. Keep it tight.
- 2–4 tags. Lowercase, hyphenated if needed.
- Fork the repository.
- Pick the right category file in
categories/. If your prompt doesn't fit an existing category, open an issue first to discuss a new file. - Add your entry at the bottom of the category file using this exact format:
---
### [Short, descriptive name]
\`\`\`
full prompt text here — one line or multi-line
\`\`\`
**Produces:** A 1-2 sentence description of what the output looks like.
**Tags:** tag1, tag2, tag3- Commit with a message like
Add [prompt name] to [category]. - Open a pull request describing where you tested the prompt and what you changed or improved.
- Prompts that produce a clearly described, usable result.
- Variations on existing styles that meaningfully differ (new lighting, new color palette, new composition).
- New category proposals that cover a clear gap.
- Prompts that only work with one specific model or tool.
- Prompts that mention specific diffusion model names.
- Placeholder-filled templates where the user has to fill in most of the prompt.
- Near-duplicates of existing entries.
- Anything NSFW, illegal, or targeting real private individuals.
- Write in full sentences inside code blocks, comma-separated where natural.
- Use photography/film vocabulary: "85mm lens", "golden hour", "Rembrandt lighting", "volumetric fog", "shallow depth of field".
- Name color palettes explicitly when relevant: "warm amber and deep violet", "cool teal and magenta".
- Describe emotional intent, not just technical details. A good prompt feels like a small story.
Open an issue. We read every one.