Skip to content
Community Research


Everything worth doing here starts with a good prompt.

Prompt engineering usually depends on conversion of expected tasks to a dataset trained on a large language model. For more about models and training you can always skip ahead.

No matter the latent diffusion model or hardware you're counting on, without practice in how to string together fluent combinations of prompt components in ways that move your chosen synthesis machine like good poetry, then good luck getting great results.

Here you can find, test, and contribute to a frequently updated collection of prompts from the image synthesis research community at large.

Be sure to also make good use of prompt search engines like Lexica and Libraire, where over 15 million prompts and counting are available as open source inspiration for you to hone and share your synth poetic skills at will.