Script Intelligence · Original quantitative study
How long are screenplay scenes?
Across 344 screened feature screenplays, the median of each script’s median scene length is 0.5 modelled pages. The middle half lies between 0.375 and 0.625.
These findings describe the available scripts in this sample; they do not set an ideal scene length.
Published 4 November 2026 · Edition 1.0 · 344 eligible scripts
What a modelled page means
The engine uses fixed text widths and spacing, with 55 lines per page. Scene headings are included. Each detected scene is rounded to the nearest eighth of a modelled page, with a minimum of one eighth. These lengths are neither physical spans in the original PDF nor estimates of screen time.
We find the median scene length within each script, then compare those script medians with equal weight. A script with 300 scenes has the same weight as one with 50. The result is not the median of every scene pooled together.
Most script medians fall below one modelled page
316 of 344 scripts (91.9%) have a median scene length below one modelled page. This is a statement about script medians, not the percentage of individual scenes shorter than a page.
View data table
| Interval | Scripts |
|---|---|
| 0 to <0.125 | 0 |
| 0.125 to <0.25 | 13 |
| 0.25 to <0.375 | 64 |
| 0.375 to <0.5 | 86 |
| 0.5 to <0.625 | 54 |
| 0.625 to <0.75 | 60 |
| 0.75 to <0.875 | 25 |
| 0.875 to <1 | 14 |
| 1 to <1.125 | 6 |
| 1.125 to <1.25 | 6 |
| 1.25 to <1.375 | 3 |
| 1.375 to <1.5 | 5 |
| 1.5 or more | 8 |
| Total eligible scripts | 344 |
median 0.5; eligible n=344. Intervals include the lower bound and exclude the upper bound. The final “or more” interval, where shown, includes the grouped tail.
One equally weighted observation per script; n=344. Modelled pages use 55 lines and nearest-eighth scene rounding. Values of 1.5 or more are grouped at the right. Edition 1.0; S1-QA-3.
- Median
- 0.5modelled pages
- Middle half
- 0.375–0.625modelled pages
- 10th to 90th percentile
- 0.25–0.875modelled pages
- Observed range
- 0.125–3.75modelled pages
Mean and median answer different questions
Within all 344 eligible scripts, the mean scene length is above the median. The median within-script difference is approximately 0.39 modelled pages. Longer scenes pull the mean upwards, so the average scene and the middle scene are useful to distinguish.
This does not establish how scene length affects audience attention, writing quality, production cost or commercial success.
Scene length by primary genre
Genres overlap, and one editorial label cannot capture everything a film does. We show counts for every genre, but withhold statistics below 20 scripts. This reporting floor does not establish representativeness or statistical significance.
View data table
Scroll sideways to see all columns.
| Genre | Eligible n | Lower quartile | Median | Upper quartile |
|---|---|---|---|---|
| Action | 36 | 0.25 | 0.25 | 0.516 |
| Animation/Family | 33 | 0.375 | 0.625 | 0.875 |
| Comedy | 56 | 0.484 | 0.563 | 0.75 |
| Crime | 12 | Not shown | Not shown | Not shown |
| Drama | 87 | 0.375 | 0.375 | 0.625 |
| Fantasy | 14 | Not shown | Not shown | Not shown |
| Horror | 39 | 0.375 | 0.5 | 0.625 |
| Romance | 20 | 0.375 | 0.5 | 0.641 |
| Sci-Fi | 26 | 0.25 | 0.375 | 0.594 |
| Thriller | 21 | 0.25 | 0.375 | 0.688 |
The middle half lies between the lower and upper quartiles. Results below n=20 are not shown. These ranges describe this corpus, not a target or quality score.
Modelled pages. Dots show medians; bands show the middle half. Each n counts scripts. Cells below n=20 are withheld. Descriptive comparisons only; edition 1.0.
Sources, screening and limits
We remeasured the same 428 source PDFs used in the September v5 benchmark. One incomplete source, Saw, was excluded, leaving 427 in the corrected v6 release. This study uses 344 scripts with eligible scene measurements. The package lists all 84 exclusions from the original input; reasons can overlap.
Prior scene-family holds remain in place. We also withhold scene metrics for unresolved image content and disagreement with an independent heading-candidate count. Seventy-five flagged front-matter pages in 69 files were visually reviewed. Clearing a front-matter flag did not clear unrelated concerns.
The eligible scripts cover released features from 2000 to 2026. Public availability and awards circulation influence which drafts enter the library. This is not a random sample of produced films or submitted screenplays. TV is outside this study. Heading conventions, draft versions, extraction errors and editorial genre labels remain sources of uncertainty.
Independent numerical checks reproduce all 46,691 per-scene numeric records used here, including rounding, per-script means and medians, overall quantiles and genre results. This establishes calculation consistency, not the accuracy of every extracted scene boundary. The review is internal and is not external academic peer review.
Extraction method: 1.6. Eligibility protocol: S1-QA-3. Quantiles use linear interpolation at (n−1)×p and retain precision; the chart labels are rounded for display. Read the wider benchmark methodology.
Does the result depend on screening?
The median and middle-half range remain 0.5 and 0.375–0.625 in the high-confidence-only subset (n=335) and the subset without dialogue-family flags (n=331). 15 included script medians change under method 1.6 compared with 1.5. A stable headline does not mean every measurement stayed the same.
A fixed edition, with a clear update history
First published edition after method 1.6 correction and S1-QA-3 screening. No new source scripts. Unpublished drafts remain archived internally. Review is due by 4 February 2027, or sooner after a material source or method change. Reviews do not automatically change the data. Corrections and expanded samples receive new editions with before-and-after notes; existing data files and citations remain available.
- Edition 1.0, published 4 November 2026
- Publication record and source hashes (JSON)
- Full results and sensitivity checks (JSON)
- Frozen article text (Markdown)
- Original distribution figure: PNG · SVG
- Original genre figure: PNG · SVG
The ZIP includes the protocol, input hashes, derived numeric records, CSVs, citation files and Python reproduction scripts. It contains no screenplay text or source images.
Explore your own screenplay
Upload your screenplay to see its page, scene, location and speaking-part counts. The measurements and their export are available on the free plan, without an AI analysis. Use this research as context, not a score your draft needs to match.
Measuring requires a free account and a laptop or desktop. On a phone, we’ll email you a link for your computer. Plan storage limits apply.
Cite and reuse this research
Share a finding, check the evidence or download the data behind this release.
The PreList (2026). How long are screenplay scenes? Edition 1.0 (scene-length-2026-11-v1). https://www.theprelist.io/research/scene-length/versions/scene-length-2026-11-v1. CC BY 4.0.
Licensed CC BY 4.0: reuse the figures, charts and dataset with attribution and a link back to this page. Snapshot scene-length-2026-11-v1; every number above traces to it. Screenplay text and third-party imagery are not included in this licence.