Synthplay

Major shifts in audiovisual technology have tended to bring new control documents with them — formats shaped as much by industrial organization as by the technology itself. Early cinema worked from brief scenarios: lists of scenes and outlines of action, of varying length and complexity. By the mid-1910s, as production industrialized, the continuity script turned those outlines into an industrial blueprint — ordering shots, locations, costs, and intertitles, and formalizing the separation of conception from execution. Synchronized sound pushed dialogue, music, and effect cues into the script — gradually, and at first across several parallel documents. And after the studio system declined, the master-scene screenplay became the dominant format: fewer prescribed camera setups, more readable scene description, with detailed shot planning deferred to directors and departments.

Animation went further. The exposure sheet took the story down to the individual frame — what happens, when, on which drawing, with which mouth shape — so that many hands could execute one vision with precision.

As audiovisual production becomes orchestrated by generative models, the assumption behind the screenplay inverts: the interpreters no longer bring human judgment, embodied continuity, or institutional memory of their own, and the image is no longer only filmed or drawn — it is generated, iterated, selected, and recomposed. A new control document becomes necessary, one that preserves narrative intent while describing computational execution.

This working draft records that document as the synthplay. The name follows the lineage of the form itself: photoplay named a passing technology; screenplay named the enduring substrate and survived every technical shift since. What endures now is that the image is synthesized rather than captured. A synthplay is to generated cinema what the screenplay was to industrial cinema and the exposure sheet was to animation: the document that turns imagination into reproducible execution.

A synthplay is an instrument of record as much as an instrument of production. It exists to do three things.

• Control Standardize executionEvery beat compiles to explicit, parameterized generation units, so that a single author can direct many models with consistency.
• Measurement Record the processEvery generation is logged, approved and discarded alike. The ratio of discarded to approved generations provides an empirical proxy for the semantic error curve, tracked as models evolve.
• Provenance Establish authorshipThe approved document forms the human-authored record intended to support provenance, authorship documentation, and, where applicable, registration; the log of discarded generations supports evidence of creative selection.

Five levels. The first two are upstream documents — the Story Bible at series level, the Treatment at episode level. The last three are the internal structure of the synthplay itself, written one per episode.

• Story Bible Defines the universePersistent canon: world, characters, visual rules.
• Treatment Defines the narrativeProse progression of the story; home of macro-structural beats.
• Scene Defines the consistency domainNo longer a container of continuous time — a grouping of shared space, character, wardrobe, and light.
• Beat Defines dramatic progressionThe smallest unit of value change. The design unit, preserved intact from the classical paradigm.
• Generation Unit Executes the beatOne generated segment — video, audio, or voice — fully parameterized: model, duration, camera, seed, continuity references, and a prompt.

The prompt is an attribute of the generation unit, not the unit itself. A beat compiles to one or more generation units; the synthplay is the source, the units are its executable form.

The beat-to-unit ratio is variable by design. It compresses along two axes: as generative models extend their usable clip duration, a beat compiles to fewer units; and as iteration counts fall, each unit stabilizes toward a single generation. The hierarchy absorbs technological change — the beat is the invariant of design, the generation unit is the variable of execution. At the limit, a single generation covers an entire beat — eventually an entire scene; the hierarchy holds, and the ratio simply falls toward one. What the methodology drives down is the semantic error curve: the distance between narrative intent and generated output.

SYNTHPLAY
├── SCENE 01
├── SCENE 02
│   ├── BEAT 16
│   ├── BEAT 17
│   │   ├── GU-17.1
│   │   ├── GU-17.2
│   │   └── GU-17.3
│   └── BEAT 18
├── SCENE 03
└── ···

The Invisible Line™, a 100-episode series currently in development and production, is the first declared case study developed under this methodology. Production records every generation, approved and discarded, so that across one hundred episodes the series doubles as a longitudinal record of the semantic error curve as generative models evolve. The specification is being calibrated against that production; it will be published when the process can be shown, not merely described.