---
name: filmscene-corpus
description: "Pull a film scene's exact structure from authorized film references for a remake — units with HD times, camera setups, cut times, camera moves, subtitle lines with speakers and word times — and split it into parts ≤ 15 s that end on a cut. Part of the filmscene reference set."
---

These creative references describe source experiments with external tools. Use your own supported account and tools; this download installs no automation. Settings and timings are experimental observations, not current FlashBoards availability. Use only media and voices you have permission to use.

# filmscene-corpus · the scene as data

## Get the scene

- From footage you have permission to use, record the scene’s shots, camera setups, cut times, camera moves and dialogue. The following field names illustrate the source’s scene notes; no private corpus or extractor is supplied.
- `units[]`: `unit`, `t0/t1` (HD seconds), `setup` (A, B, C… = camera setups; same letter = same angle), `labels` (крупность, состав, ракурс, свет), `camera[]` moves with times.
- `cuts[]`: the cut times, J/L cuts (`audio[]` — a line that crosses the cut).
- `lines.subtitles.rows[]`: `t0/t1`, `who`, `text`, `units`. A subtitle row can hold two speakers — split it by the diarized STT of the original (Scribe v2, `speaker_id`).
- `who` is the film's role. Map each role to the person's cast ({character A} replaces <film role 1>, {character B} replaces <film role 2>; ask the person if unclear) and carry that mapping into setups, `speakers` and voices.
- Cut the scene once from your authorized HD copy and work relative to its start.

## Parts

- **≤ 15 s, end on a cut.** Sum units from the start; if a part runs a little over (15.14 s), trim the head before the first line (0.14 s) rather than the tail.
- The generation length is a whole number of seconds; when the part is shorter (12.888 s), generate 13 s and trim in the edit.

## Acting

Make a 1 fps contact sheet of the part (`fps=1,scale=400:-1,tile=4x4`) and write what each person does — gaze, head, mouth, nods — shot by shot. It goes into `action` and `acting` of the prompt.

## Example run (reference only — never reuse its files, prompts, faces, voices or frames for a person)

*Heat* (1995), scene 95 from unit 996: part 1 = 996_0 (A) + 997_0 (C) + 997_1 (A) = 15.0 s (the 15.14 s case); part 2 = 998_0 (C) + 998_1 (A) = 12.888 s; part 3 = 998_2 (C, 12.69 s)… A two-speaker subtitle row: "Enough time to do what you wanna do? That's right." Scene file `orig/heat_original_scene.mp4`.

That filename and those corpus unit labels identify the historical experiment only; the file and corpus are not delivered or required. Substitute your own scene file and measured shot labels.