TheFluxTrain

TheFluxTrain is an agentic film studio that converts story or promo briefs into scenes on an editable timeline, using template-driven node workflows, while the company researches a persistent world model for long-form film.

TheFluxTrain is a film production platform built around an agentic studio. A user supplies a story or promo brief, optionally with up to nine reference images, and the system builds scenes, wires a generation pipeline, and lands the result on a timeline where shots can be trimmed, subtitled and exported. The company positions the studio as a working product today and describes a longer-term research effort: a world model for film that keeps characters, locations, wardrobe and camera state persistent across a whole production instead of rediscovering them in every clip. The pitch rests on a frustration with current clip generators, which tend to drift on character identity, struggle to join clips into a fluent sequence, and force a full regeneration when a single beat is wrong. TheFluxTrain's planned answer is to predict the next short state from earlier frames, references and a director instruction, which would allow local fixes without redoing a shot. That model is still research, while the studio is what can be used now. In practice the studio is a node-based workflow editor. Templates load pre-wired graphs for things like multi-style Seedance 2.0 text-to-video ads, clay-figure or anime character renders, product posters, beverage campaigns, photo restyling, and a video restyle flow that swaps the look of a person while keeping the original motion. Each node has its own generation controls, output sinks for picking the best variant, and iteration carousels. Tab-group nodes fold repeated subgraphs together, which suits teams producing ad variants or product SKUs. A tool called Director Zero handles story-to-scenes generation, and a command line workflow for file uploads is referenced in the tutorials. The audience is split into AI-first agencies running client pipelines and batch variants, creators making series and short-form films, and parents producing family story videos from photos. Compared with single-prompt video generators, it trades simplicity for a graph-based, multi-model workflow that spans images, video, voice and editing, and it consumes credits per generation node.