Directed Generative Production (DGP) Methodology: A Control Architecture for Generative Video

Generative video systems can produce visually sophisticated sequences, but compelling imagery is not equivalent to executing a specified production. A renderer may generate plausible actions, environments, and camera movements while departing from requirements concerning spatial relationships, character identity, chronology, object continuity, framing, and editorial structure. This paper introduces Directed Generative Production (DGP) Methodology, a requirements-driven framework developed by Jeffrey Seders for planning, executing, qualifying, and verifying generative video productions. The methodology incorporates visual references, first-frame and last-frame constraints, spatial blocking, shot/action chronology, continuity controls, renderer qualification, and post-generation verification. It distinguishes visual plausibility from production compliance and emphasizes measurable evaluation against predefined requirements. Developed through applied research beginning in 2026, DGP Methodology has been investigated through cross-renderer comparisons, character-continuity experiments, single-generation editorial sequences, and longer-form narrative productions. The central proposition is that generative video becomes more controllable when production requirements are explicitly represented, assigned to appropriate information channels, and independently verified. Author: Jeffrey Seders, Independent Researcher. Research website: https://www.jseders.net