Most prompt libraries ask you to trust a caption. Promptweet started from a simpler idea: show the clip next to the text that made it, and let the two speak for themselves. If a prompt says “handheld, low-angle, volumetric light” you should be able to see exactly what that combination looks like before you copy a single word.
What you are looking at is that idea built out: a working interface over a real, growing corpus, pulling actual prompts posted by real creators so you can browse by camera move, lighting or duration and see exactly what each choice looks like.
Every claim on this site, whether it is a rank, a percentage or a “most used” label, is computed live from the 120 prompts in the library right now, across 26 categories and 6 model libraries. Nothing is asserted from memory or guessed; the corpus keeps growing and every number on the site recalculates with it.
A small crew: one person who reads too many prompts too closely, one who draws the interface around what the data actually shows, and one whose only job is to catch a percentage that does not add up. That is the whole team, and it is enough for a library this size.
The interface, the filtering logic and the corpus statistics are genuinely computed. Most prompts, handles, dates and clips are pulled from real posts on X — each one links back to its source, so you can go verify it yourself. The library is still filling in: a handful of models don't have real prompts yet and show placeholder content in the meantime, marked as sample data wherever it appears.
Questions, feedback, or just want to talk about camera moves, the contact page is the way in.
