How do you create a shot breakdown for an AI film before you start generating video?

Build the shot breakdown inside an agent that holds your full script: load the screenplay into a creative producer agent, lock characters, props, and deliverable format, generate a scene-by-scene shot list, then spec every shot with the fields AI generation needs — prompt, negative prompt, references, length — and sequence the list before generating anything.

Start by loading the complete screenplay into a creative producer agent — this gives the breakdown full narrative context (characters, arcs, themes) so every downstream shot decision is grounded in the same understanding. invideo is an agentic video creation tool with all the current generation models available, so you build the breakdown and generate from it inside one context. In one documented multi-agent production, the creative producer agent held the script, shot breakdown, and character details as the foundation every other agent referenced.

Before the list gets written, answer the foundational questions that change every frame. In one documented session the invideo agent asked four things before building any assets: what the main character looks like, what the antagonist references, what the key prop is, and the deliverable format. Lock those answers — plus your character sheets and environment references — first, because every shot in the breakdown will point back to them.

Next, have the invideo agent generate a scene-by-scene shot list directly from the prose script. The agent structures and numbers it itself; one large production's notebook ran to scene 169, with shot variants numbered 21.1–21.5 under a single scene. If you have a style or treatment document, load it before this step so every shot in the list inherits the visual grammar automatically.

Then spec each shot for AI generation, not just framing. This is where an AI shot breakdown diverges from a traditional shot list: in one documented production the invideo agent output 12 parameters per shot — film reference, shot design, length, style interpretation, emotional register, lens, lighting plan, color script, atmosphere layers, blocking, the final prompt, the negative prompt — plus a revision prompt for retries. The generation prompt is a field in the breakdown, not something you improvise later. Set each shot's length around your video model's generation unit — Seedance 2.0, for example, returns multi-shot clips, so one breakdown entry can cover several shots rather than a frame-by-frame storyboard.

Sequence the breakdown before generation begins. Run a director's assistant agent specifically to tighten shot order — it should know which shot comes after which and how the edit flows before any video is executed. A storyboard agent visualizing each shot first gives you a visual brief that makes the per-shot direction more precise.

Finally, validate the breakdown against model limits before spending credits. The invideo agent can flag shots that exceed what a model can deliver: in one production it caught a scene scripted at 18 cuts in 15 seconds and recommended splitting it into two parts — which produced a sharper result than the original script. For longer films, work act-by-act: divide the script into acts and fully complete the breakdown and generation for one act before starting the next, which prevents the invideo agent from losing context on long-form projects.

Watch some of these to see what works for you:

To really set up the context for the agent, I normally start off with the creative producer agent. That's where I'll give the script, or the shot breakdown, along with the characters. That's the main agent that sort of holds the understanding and the vision of the entire film.

— invideo filmmaker, from a documented multi-agent production