AI Episode Maker

What happens inside one AI-made episode: units, redos, versions and the cut

The episode register as a data file

Every logged statement, 171 in all, is in one CSV file with its editor, behaviour, source page and dates. It is released under ODC-By 1.0. As of 2026-09-25.

One statement per rowA row holds a single logged statement: which editor and behaviour it belongs to, the reading as a sentence, where it sits on the vendor's page, the page address and the dates.A rowfact_idEditor, behaviour, number.valueThe reading, as a sentence.sourcePage address and title.datesPage date if given; day read.
Fig. 1 The same rows drive the editor pages and the behaviour pages.
Columns in in-episode-register.csv.
ColumnHolds
fact_idEditor, behaviour and a number
valueThe reading as a sentence
applies_toWhere on the page the statement sits
source_urlThe page it was read from
source_nameThe page's owner and title
value_sinceThe date the page gives itself, if any
verifiedThe day it was read

Inclusion rule. Every column in the file. Order. File order.

1What the file is for

Sorting the file by behaviour shows every editor's wording on, say, redo charges in one column. Sorting by editor gives one tool's full record. Behaviours a vendor skips are included as rows with the file's empty wording, so a count of silences is part of the data.

Readings that belong to a closed list, such as who controls transitions, are in the value column as sentences; the short labels used on the register are derived from them and are not stored separately.

The applies_to column is the one to read before quoting a row. It says which part of the vendor's page the statement comes from, a FAQ answer, a plan note, a step in a tutorial, so a reader can find the sentence again even if the page has been rearranged since it was read.

The file carries no prices. Redo readings record the basis of a charge, never an amount, because amounts change more often than the rules behind them.

Download in-episode-register.csv

2Sections

3Inside a generator's episode

The episode pipeline SceneMixer documents has its own answer to “Data”; two lines from its pages:

Paste a scene or a short script and SceneMixer gives you a storyboard — numbered shots with framing and seconds — and then a moving version of each segment, which you can join into a rough cut of the scene.

SceneMixer, Using SceneMixer for Previs and Animatics Before a Live-Action Shoot (read 2026-10-13)

Typically 60–120 seconds, always 9:16 vertical, ending on a hook. SceneMixer's episode outlines are generated with that pacing, and setting a project to 9:16 makes every segment vertical (the default is 16:9).

SceneMixer, How to Make ReelShort-Style Dramas with AI — “How long is a ReelShort-style episode?” (read 2026-10-13)

The quote column. Each reading in the file also carries the words on its source page that it was read from, so any row can be checked against that page. Rows recorded as not stated on the pages read leave it empty.