How Shutap is built
Shutap writes a draft from your own story, then hands it back for you to steer. Twelve observations come first, four whole bits are written from different ones, rules written in code filter them, a second AI model ranks what is left, and you can make the winner hotter, tighter or more escalated, or rebuild it from lines Shutap wrote for your story.
Turn your own story into a bit
Write what happened. Get a bit: hook, setup, tags and button.
What makes it different
Most joke generators take a topic and return a line you either use or do not. Shutap works the other way. It starts from your story, writes a whole bit with a hook, setup, tags and a button, and then gives you controls to change it. You stay the writer. Shutap is the draft and the sparring partner.
You start with a story, not a prompt
You write what happened, the way you would tell a friend. Before anything is saved, names, places and contact details are replaced with role labels, and you can see what changed. If a story reads as heavy, Shutap does not turn it into a joke. It stops and points to real help instead.
Step one: find what is actually funny
Shutap does not begin by writing jokes. It first lists twelve observations about your story: what is strange about it, what someone said against what they did, what a word was hiding. The list is saved with the story, so every rewrite later starts from the same observations instead of guessing again.
Step two: write whole bits, not lines
Four complete bits are written in one pass, each built on a different observation, plus five spare tags. Every bit has a hook of eight words or fewer that can sit on screen for two seconds, a setup that states the facts plainly, one to three tags that each end on something you can picture, and a button. Four controls shape the writing: audience (social, live, office, work, family or school), voice (deadpan, storyteller, roast or dry), length (15 seconds to 2 minutes) and how wild, from 1 (true to your story) to 5 (goes wild). It sets how far the jokes stray from the facts, never who they're aimed at.
Step three: rules in code, then a judge
Rules written in code run first, so they hold even when a model slips. No advice. Jokes go at the situation and the other party's behavior, never at the teller or a real person. No invented facts about someone else. No jokes about bodies, age, looks, intelligence, race, religion, sexuality or disability. And the swap test: if a tag would work just as well on a stranger's story, it is cut. The bits that pass are ranked by a second AI model from a different family than the writer, so the model that wrote a bit never grades its own work. It ranks on whether there is a picture, whether the hook would stop a thumb, whether the button lands instead of explaining, and whether a person would actually say the line out loud. If the judge is unavailable, the first bit that passes the rules is used.
A library of comedy technique
Behind the prompts is a library of named comedy techniques: how a joke gets a picture, how tags escalate on one premise, how a button turns a situation around. There is also a hall of fame of lines the founder approved as funny. It sets the bar for what counts, and new writing is checked against it so Shutap does not repeat those lines. The library teaches Shutap how funny writing is built. The writing you get back is about your story.
A dataset we write ourselves
The writing is guided by an in-house dataset the Shutap team writes, selects and rates, pairing everyday situations with jokes, tagged by technique, voice and how wild. None of it comes from users. It grows as the team adds to it, so the guidance gets better over time.
What to post today
Never start from a blank page. The Shutap homepage shows a daily idea of what to post, with a hook, a short bit and a tag, drawn from what is trending, alongside a trend panel that shows which topics are rising or cooling.
Learn from the funniest
The funniest is a series of educational pages. Each asks who is the funniest in a category, such as US presidents, using published rankings and sources, then breaks down how that person was funny into techniques you can use, each with an original example line and a way to try it. Read them at shutap.com/funniest.
You steer
Hotter takes it two levels further from the facts, up to 5. Tighter cuts the bit to about 60 percent of its words and keeps only the best tag. Escalate moves each tag one step further along the same line and flips who is in charge on the button. Swap a tag trades one line for another. Mix your own lets you pick the hook, the tag and the button you like from the lines Shutap actually wrote for your story, and rebuilds the bit around the setup. Every change is saved as a new version, so you can go back.
Then film it
Each bit can run in a full-screen teleprompter, as a scene laid out beat by beat with shots, speakers and on-screen text, or as a screenplay page with scene headings, action, character names and dialogue. Length is estimated from the word count at about 2.6 words a second.
Versioned, tested, and honest about limits
Every bit is stamped with the version of the prompts that wrote it. Changes to the writing are checked against a fixed set of test stories that are never used as examples. Shutap is written by AI and gets things wrong, so you decide what to perform. Test it on a real audience before you trust it.
Write your story, get the bit →
Turn your own story into a bit
Write what happened. Get a bit: hook, setup, tags and button.
