53%
One-turn speakers in the largest-cast quartile
Up from 26% in the smallest-cast quartile.
Original screenplay research · 994 scripts
The median screenplay in this corpus has 47 speaking characters, but a much smaller group carries the dialogue: the same distribution works out to an effective 6.3-speaker core.
What is more telling is how that gap widens. As casts grow, scripts add far more one-line, few-line and single-scene speakers than central voices — the edge of the cast expands much faster than the conversation at its center.
Produced screenplay research
Look at how widely dialogue is shared across produced feature scripts. Compare speaking-cast size, the largest speaker’s share, and the balance of dialogue and action lines.
Source: The PreList Industry Intelligence (2026). The Anatomy of a Produced Screenplay.
Detected character cues include minor speakers, not just principal roles. This measurement has a smaller eligible sample than page counts.
All eligible scripts · 233 eligible scripts
detected speakers
Tallest bar: 56 scripts
| Interval | Scripts |
|---|---|
| 0 to <10 | 3 |
| 10 to <20 | 16 |
| 20 to <30 | 34 |
| 30 to <40 | 55 |
| 40 to <50 | 56 |
| 50 to <60 | 30 |
| 60 to <70 | 15 |
| 70 to <80 | 11 |
| 80 to <90 | 6 |
| 90 to <100 | 3 |
| 100 to <110 | 1 |
| 110 or more | 3 |
Choose a genre to compare its median and middle half with the complete eligible sample.
Use cast size to plan an introduction pass, and the largest-speaker share to inspect who drives exchanges. A large share may be the point of a protagonist or a monologue; the number does not decide whether it works.
Produced feature screenplays, 2000–2026. Eligible samples differ by measure. Genre groups follow the publisher’s classification. These are selected produced scripts, not a representative sample of specs or a measure of quality.
Inkwell extracted the published aggregates and created these comparisons and charts. No underlying screenplay files were remeasured.
Versioned source · Measurement definitions · CC BY 4.0
Download the measurements, attribution and source notes (JSON)Try it on the page
A few longer speeches can carry a large part of a scene’s dialogue. Compare their share of turns with their share of words, then use that distinction to choose where to start a revision pass.
Pooling the sampled turns, turns longer than 23 words account for 9.5% of turns and 34.6% of words. On a dialogue pass, inspect those longer speeches for a change in tactic or new information. Length alone is no reason to cut one.
Across films, the middle 80% of word shares above this threshold runs from 17.2% to 47.8%; the median is 31.8%. These are differences between the sampled films, not a recommended range for your script.
This Cornell-derived numeric projection contains 304,439 word-bearing sampled utterances across 617 films; 274 zero-word rows are excluded. The Cornell sections below examine exchanges and credited cast from the same source corpus. These are sampled utterances, not complete screenplays. Thresholds are exclusive: “longer than 23” starts at 24 words.
Pooling gives more weight to films with more sampled turns or words. Equal-film weighting averages the within-film percentages. Film percentiles use linear interpolation. Neither view measures performance time or writing quality.
Download the measurements and methods (JSON)Next, inspect how exchanges unfold or who carries the dialogue.
Try it on the page
The percentage can change because the denominator changed. Compare three counts of the same excerpt before treating a dialogue percentage as a revision note.
One original excerpt, explicitly labeled
| Count by | Dialogue / total | Share |
|---|---|---|
| Words | 18 / 32 | 56.3% |
| Paragraphs | 2 / 3 | 66.7% |
| Nonblank lines | 2 / 5 | 40.0% |
The words and the scene stay the same when you join the action lines. Only the line-based percentage changes. Before comparing two “dialogue percentages,” check what each one counts.
Words are whitespace-separated tokens. Each labeled block is one paragraph; lines are explicit nonblank text lines, not visual browser wrapping. Headings and character cues are outside this example.
A dialogue-to-other-text ratio of 3:2 means a dialogue share of 3 ÷ (3 + 2) = 60%, not 150%. These counts describe this excerpt, not an ideal mix.
The strongest signal
We measured the edge of the cast directly: speakers with a single dialogue turn, speakers with five or fewer, and speakers who appear in only one dialogue scene. All three become far more common as casts grow.
Who lives at the edge
Median share of speaking characters meeting each definition in the smallest- and largest-cast quartiles.
The median speaker receives 5 turns in the smallest-cast quartile and 1 in the largest.
53%
Up from 26% in the smallest-cast quartile.
79%
Up from 53% in the smallest-cast quartile.
69%
Up from 47% in the smallest-cast quartile.
The pattern holds across the whole sample, not only at the extremes: the larger the cast, the higher its share of one-turn, few-turn and one-scene speakers.
Raw cast versus conversational core
In the smallest-cast quartile, the median screenplay has 27 distinct speaking characters and a 5.1-person effective core. In the largest, the raw cast rises to 90 while the effective core reaches only about 8.0.
Cast size versus speaking core
Scripts are grouped by speaking-cast quartile. The pale bar is the median number of distinct speaking characters; the darker bar shows the effective speaking cast after accounting for how unevenly dialogue turns are distributed.
Cast quartile 1
n=256
27 raw · 5.1 effective
Cast quartile 2
n=258
41 raw · 6.3 effective
Cast quartile 3
n=237
55 raw · 6.5 effective
Cast quartile 4
n=243
90 raw · 8.0 effective
| Quartile | Scripts | Median raw cast | Median effective cast | Effective share |
|---|---|---|---|---|
| 1 | 256 | 27 | 5.1 | 20.4% |
| 2 | 258 | 41 | 6.3 | 15.5% |
| 3 | 237 | 55 | 6.5 | 11.7% |
| 4 | 243 | 90 | 8.0 | 8.2% |
The effective speaking cast just makes unequal speaking loads visible. If dialogue were spread evenly it would match the raw count; the more a few characters dominate the turns, the smaller it gets.
Who carries the turns
The most frequent speaker carries a median 30.6% of turns, and the top three carry 58.6%. That concentration is why a long cast list can still read like a story built around a small recurring conversation.
Median share of dialogue turns carried by the most frequent speakers in each screenplay.
Scene shape
Give each screenplay equal weight and about 26% of dialogue-bearing scenes have one speaker, 37% have exactly two, and 37% have three or more — three shares of one whole, so they total roughly 100%.
One speaker
26%
Two speakers
37%
Three or more
37%
A second corpus · Cornell Movie-Dialogs
The Cornell Movie-Dialogs Corpus — 617 films, 304,439 word-bearing utterances, 3,210,011 words — measured the way the scripts above were, one film at a time with each film weighted equally: the median film's median utterance is 7 words, the same as the median script's median dialogue turn (7). The middle 80% of films sit between 5 and 9 words.
7 words
The 90th percentile is 23 words. Most sampled dialogue is short.
24%
Those long utterances are only 5.4% of the sample, yet with each film weighted equally they hold 22% of the words.
36%
Short is the norm; the long turns are where the words are.
Inside an exchange
Cornell records 83,097 conversations, each an ordered run of turns between two characters. The median exchange is 3 turns (mean 3.7, longest 89); 46% are exactly two, and 99% of multi-turn exchanges strictly alternate speakers. The median exchange holds 26 words; one in ten runs past 82.
57%
In a typical exchange, one turn carries more than half the words.
48%
Nearly half of all exchanges are one character talking and the other replying.
3 turns
Concentration at the level a writer revises: the exchange, not the whole script.
Share of exchanges by the credited genders of the two characters, among the 52,038 exchanges where both are labeled (63% of exchanges, 66% of words).
| Pair | Share of labeled exchanges | Share of their words |
|---|---|---|
| Two women | 6.0% | 5.8% |
| A woman and a man | 49.2% | 48.5% |
| Two men | 44.8% | 45.7% |
Credit position and credited gender
The corpus labels 3,015 of its 9,035 characters with a gender and 2,679 with a credit position, both taken by its authors from cast lists — two cohorts, each about a third of the cast. The gender-labeled characters speak 79% of the corpus's words and the credit-positioned characters 70%, so what follows describes the principal cast and names the cohort behind each figure.
Words by credit position
Share of all words spoken by credit-positioned characters, grouped by where the character sits on the film's credits; each bar is that share of the full track. The right column is the median words one character in that group speaks, and the share of the group's words spoken by women.
Top billing
n=497
35% of words · median 1,431 · women 17%
Second billing
n=443
22% of words · median 906 · women 47%
Third billing
n=352
12% of words · median 650 · women 32%
Fourth to fifth
n=479
12% of words · median 483 · women 26%
Sixth and later
n=908
19% of words · median 337 · women 29%
| Credit position | Characters | Share of positioned words | Median words per character | Women's share of the group's words |
|---|---|---|---|---|
| Top billing | 497 | 35.4% | 1431 | 16.9% |
| Second billing | 443 | 21.6% | 906 | 47.0% |
| Third billing | 352 | 12.1% | 650 | 31.5% |
| Fourth to fifth | 479 | 12.4% | 483 | 26.1% |
| Sixth and later | 908 | 18.6% | 337 | 29.4% |
Among labeled characters, women are 32% of the cast and speak 28% of the words. In the 519 films with both labels, the median film gives women 27% of its labeled words, and in 17% of films women speak the majority. 89 of the 497 top-billed characters are women; a top-billed woman holds a median 36% of her film's labeled words, a top-billed man 40%.
28%
Women's share of labeled words
27%
Median film: women's share
17%
Films where women speak the majority
18%
Top-billed characters who are women
Put it against your own draft
The free dialogue-balance tool breaks down your own Fountain or Final Draft screenplay: who carries the turns, the effective cast size, the scene mix.
About the data
This analysis uses Movie Script Database — IMSDb tagged archive, published by Aveek Saha on Zenodo under CC BY 4.0. Of 1,088 tagged files, 994 met the dialogue and scene thresholds used here.
The replication, exchange and credited-cast sections use the official Cornell Movie-Dialogs Corpus release by Cristian Danescu-Niculescu-Mizil and Lillian Lee: 617 films, 9,035 characters, 304,713 utterances and 83,097 conversations.
These are corpus descriptions, not screenplay rules. Both archives are convenience samples, and the IMSDb reading is imperfect. Use the distributions to make a vague structural question specific, not to force a draft toward an average.
A dialogue turn is one block of speech beneath a character cue. The effective speaking cast weights frequent voices more heavily than characters who speak once or twice, then reports that concentration as the number of equally active speakers that would produce the same pattern.
The IMSDb analysis was last run 2026-08-21; the Cornell analysis on 2026-09-18.
What the edge-heavy cast pattern means for a rewrite.
Read nextSeven-word turns, longer speeches, and what dialogue length can actually tell you.
Read nextWhere action, scenes and cast arrivals land across the draft clock.
Read nextPlot keeps screenplay structure, revisions and agentic writing tools in the same workspace.