Comic Critique.
The one person who cannot hear a set properly is the person who was on stage doing it. Paste the transcript of a tape and get back what the room heard and you did not — your pace, the phrase you lean on, the swearing under your breath, where you stopped talking — and the ums, where your captions kept them. It runs on this device. Nothing is uploaded.
You already have the transcript
YouTube writes one for every video it captions, unlisted ones included. Three steps, on a computer.
Read a set
What the tape says
Set to set
One tape tells you about one night. Keep the numbers from each read and see whether they move — fewer ums a minute is a thing you can actually watch happen.
Full specials, read the same way
These are full specials on YouTube, each run through the read above exactly as a pasted tape is — the same counts and the same rules — so a finished hour sits beside your own five minutes as a yardstick. A word cloud, the topics and a plain description of the style come with each. Read on 1 October 2026.
Side by side press a column to sort it — a higher number is a different style, not a better one
- The numbers are the read-out's own: words a minute over the whole set, laughs included; filler a minute (um, uh, ah); swearing a minute, with a censored word on automatic captions counted; the longest silence, the time after a line beyond what its words take to say.
- Language is the read-out's band: clean when no swearing is on the captions, some language under 1.5 words a minute, blue from 1.5 a minute.
- Storyteller, setups that build, laughs close together come from the words between the laughs the captions mark — 80 and more, 41 to 79, 40 and under — and only where the captions mark at least one laugh a minute.
- Brisk or unhurried pace: 170 words a minute and up, or 110 and under.
- Act-outs: a voice set up and then played — “I'm like”, “she goes”, “he said” — two a minute and more. Crowd work: someone in the room captioned speaking, or asked a question from the stage, four times or more, or a stretch of the set spent with the audience.
- Topics come from each chapter's title, or where it names none, from the words the chapter uses most, weighted by the minutes it runs. The word cloud is the read-out's own count of the words that carry a topic, with the comic's own name and any rough word left out.
What this does not do
It cannot tell you whether anything was funny. A transcript sees the words, not the face, the timing or the room — so where a question needs a person, the read-out says so instead of guessing.
Auto-captions are a rough draft. They drop ums, mishear names and almost never mark a laugh, so treat every count from them as a floor. The clearer the sound on the tape, the fewer words they mishear — Stage Time → Tech shows how to tape a set you can hear.
Nothing leaves this device. The transcript is read inside this page and gone when you leave it; only the summary numbers are kept, and only if you press Keep.