Labs · Prose stats
A passage’s build:
rhythm, dialogue, repeats.
Paste a chapter. Get the sentence-length histogram, the variation number our model evaluations lead with, dialogue density against real-corpus bands, and any verbatim repeats — structure and rhythm, not spelling and synonyms.
Ctrl/Cmd + Enter to analyze · text is processed locally, nothing uploads.
The prose report appears here.
Where the numbers come from
The metrics are the structural half of our fiction-continuation evaluation line: the same sentence-splitting and dialogue rules that scored nine models over twenty continuation rounds each, ported to the browser and cross-checked against the original implementation to nine decimal places in tests. Sentence variation is population standard deviation over mean; dialogue counts quoted spans (both curly-quote families and paired straight quotes) against the full passage; repeats are verbatim segments of sixteen or more characters found by sliding-window fingerprinting.
The reference bands were computed 2026-07-24 from 378 chapter samples in 10 in-house novels — corpus we own outright, which is also why the sample button pastes an original passage written for this page rather than anyone's copyrighted work.
What the numbers can't see
These stats describe texture, not quality — plenty of great chapters sit outside every middle band. Verbatim-repeat detection catches literal loops (the failure mode where a model re-types the same paragraph); thematic repetition — saying the same thing in fresh words — passes right through. And structure is only one layer of the AI-tell problem: our double-blind experiments found evaluator LLMs judging human writing as AI at embarrassing rates, so treat any single automated signal, including this one, as a lens rather than a verdict.
FAQ
What is the sentence-length variation number, and why does it matter?
It's the coefficient of variation — standard deviation of sentence lengths divided by their mean. Human fiction mixes three-word jabs with forty-word sprawls; machine continuation tends toward the middle. In our nine-model, twenty-round continuation experiment, human originals sat around 0.64 while bare-template AI drifted down to 0.52 round after round. Uniformity is the first structural tell, which is why this page leads with it.
Why doesn't this page give my text a score?
Because the honest output is a description, not a grade. A low dialogue ratio doesn't make a book worse; a jumpy rhythm isn't automatically 'better prose'. The reference bands tell you where a passage sits among 378 real chapter samples — what you do with that position is a reader's judgment, and we'd rather leave it one.
How is this different from the AI-Flavor Checkup?
Different layer. The AI-Flavor Checkup reads vocabulary — stock phrases and sentence molds ('a flicker in the eyes', 'a hint of'). This page reads the skeleton: length distribution, dialogue share, paragraph shape, verbatim repeats. A text can pass one and fail the other; running both gives you the word-level and structure-level views of the same passage.
Where do the reference bands come from?
From 378 chapter samples across 10 in-house Chinese novels, computed 2026-07-24 with the same sentence and dialogue rules this page runs. In that corpus, sentence variation's middle band (P25–P75) is 0.62–0.72 and dialogue ratio's is 22–40%. They're reference points for 'where does my text sit', not quality thresholds.