How do you run casting and world-building at the same time using AI agents to speed up film pre-production?

Run casting and world-building in parallel by initializing a creative producer agent with the full script first, then branching a casting agent and a world-building agent on separate project pages working simultaneously. In one documented production, 3 people plus one AI agent locked cast, costumes, look and feel, and world images in a single day.

invideo is an agentic video creation platform with all the current image and video models available, so one team can run several specialized agents at once inside it.

1. Initialize a creative producer agent first. Give it the full script, shot breakdown, and character details before you branch anything — this agent holds the vision of the whole film and grounds every agent you spin up afterward in the same creative understanding. Tell it upfront how you want to work: what assets you'll share next and what it should ask for. Before it builds assets, have it surface the foundational questions that change every frame — what each character looks like, any antagonist reference, prop specification, and your deliverable format — and answer them once so both tracks inherit the same answers.

2. Branch a casting agent on its own project page. Instruct it to run the same character prompt on two image models simultaneously rather than sequentially — for example Recraft, which renders pores, lines, and stubble that make faces read as real, against GPT-Image-2 — then pick the aesthetic you prefer and develop it into multi-angle character sheets (Nano Banana handles these well). Include close-up panels, not just wides, so small details like scars and accessories survive across models. In one documented production, locking one character took about 5 generations at ~$9.78 per character. For costumes, mood-based direction works when you have no exact spec: describe the feel of the character and the invideo agent returns multiple concrete options to select from.

3. Branch a world-building agent at the same time, on a separate page. Batch your visual references by theme — spatial logic, screen function, color theory — and feed each batch with explicit instructions on what to adopt and what to ignore. Have it generate image grids instead of single frames (one documented workflow requested 3 grids per round), iterate on the grids you like, then extract the best panels; those extracted panels replace your original references and become continuity anchors for every scene that follows. Once you lock one world element, the invideo agent extracts every angle — wide, close, side — without you requesting each individually. Keep anchor images on a dedicated project page, separate from the main world-building page.

4. Keep the tracks isolated, then merge through the producer agent. Separate project pages are the anti-contamination mechanism: you give targeted feedback to the casting agent without polluting the world-building agent's context, and vice versa. When both tracks lock, the producer agent holds the merged state — locked character sheets plus world anchors become the generation seeds for production, which produces stronger shot continuity than reference images alone.

The time compression is documented. Day 1 of a 5-day production sprint ended with cast locked, costumes locked, look and feel locked, and world images generated — by 3 humans and one AI agent. One filmmaker runs 6 agents simultaneously as a standard setup; another deployed 8 specialist agents across separate project pages to finish a 2-minute brand film in 3 days, versus an estimated 1 week of manual prompting or roughly 2 months of traditional production. Practitioners credit the parallelism, not just the automation: the advantage is running many iterations at once, which is what makes the 5x-faster pipeline claim hold up in practice.

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.

— a filmmaker documenting a multi-agent short film production on invideo