Genspark puts an agent director in charge of video: one sentence of need, then script, storyboard and final cut on one canvas
Synopsis
Genspark has launched Media Studio, an AI studio that moves video production from a pipeline of script, cast, sets, voice, score and editor into one workspace run by an agent director. A user describes the piece and answers a few questions; the agent settles concept, script, storyboard and budget first, then orchestrates image, video, audio and 3D generation models from more than a dozen providers to generate, check quality, edit and composite. Output ranges from a 15-second ad to a 10-minute short drama, and from a poster to an interactive film.
Interpretation
Video production moves from a sequence that needs a script, a cast, sets, a voice, a score and an editor to a task that starts with describing what you want. Before, most people with the idea stopped at the first item on that list, and those who kept going spent weeks learning a workflow built for professionals. Genspark draws the difference at the workflow: the video tools people have already tried still demand a professional workflow and years of editing experience, while here the entry is a description and a few questions, and the studio carries the production.
A single generation model does not make the video; an agent directs it. That director's planning, script and code come from two language models, Claude Opus 5.5 and Fable 5.1, while GPT-6 Astra runs multimodal verification in the 3D director stage, checking generated results across modalities, and the agent then orchestrates dozens of generation models across image, video, audio and 3D from more than a dozen providers, among them Seedance 2.5, GPT Image, Nano Banana, ElevenLabs and Seed Audio. In operation, the agent agrees on concept, script, storyboard and budget before rendering anything, then runs generation, quality checks, editing and compositing, returning each key checkpoint to the conversation, where a frame can be circled and annotated.
Perspective
It is aimed at people who want to make the piece from the idea: one entry point and one pipeline, with script, characters, sets, storyboards, 3D blockout animation, voiceover, music and sound effects on one infinite canvas, after the first cut the assets pass to a layer editor, a multi-track timeline and a 3D white-model director stage, and agents can call the same tools.
Genspark gives the applicable range by example, a 15-second ad at one end and a 10-minute short drama at the other, which is the product's own stated boundary; the model roster changes as new models ship, so the available generation models are not a fixed set.