MuchSearch · AI Prompts

AI Prompt Search — PromptHero, Lexica, Civitai

Search PromptHero, Lexica, Civitai and FlowGPT for image and chat prompts by subject. One input, several libraries.

5 search sourcesType once, search manyFree, no sign-up

All sources

AI Prompts sources

Open a source directly, or type keywords and search it right here.

PromptHero

Millions of image prompts, searchable by subject

A large image-prompt library organised by model and style, with the generated image beside each prompt. Best for learning the vocabulary that produces a specific look.

Lexica

Stable Diffusion prompt and image search

Stable Diffusion prompt and image pairs in a fast visual grid. Search by subject and copy the prompt structure rather than the exact words.

Civitai

Open model and image prompt community

Community hub for models, LoRAs and the prompts that produced them. Essential when you need to match a prompt to a specific checkpoint.

FlowGPT

Chat prompt library shared by the community

User-shared chat prompt flows and characters, organised by use case. Quality varies widely — read the example output before trusting a prompt.

Awesome Prompts

GitHub prompt collections via code search

The GitHub collection of role-based chat prompts, searchable through code search. Plain text, versioned and easy to fork into your own notes.

About the AI Prompts channel

Writing a prompt from scratch is slower than finding one that already works. Prompt libraries are the shortcut: thousands of image prompts with their output attached, and chat prompts written for specific jobs — a code reviewer, a cover-letter editor, a market-research analyst.

Search by subject, not by tool. “Cyberpunk street market at night” will return usable image prompts across libraries, and you can then adapt the phrasing for the model you actually use. For chat prompts, search the task (“summarise a legal contract”) rather than the model name — good prompts survive model upgrades.

AI Prompts search — frequently asked questions

Are prompts from these libraries free to use?

Prompts themselves are generally free to copy and adapt. The images beside them belong to whoever generated them, so do not reuse the output commercially without checking that library’s terms.

Why does a copied prompt produce different results for me?

Prompts are model-specific. A prompt tuned for one checkpoint or model version will drift on another; keep the structure and adjust model-specific tokens.

Is it better to search by model name or by subject?

Search by subject first, then narrow by model inside the filters of each library. Subject searches return prompts from several models, which is how you learn which phrasing is model-specific and which is universal.

Where should I start if I have never written a prompt?

Pick one image you like on PromptHero or Lexica, copy its prompt, then change a single element at a time. That teaches more than reading a guide.