Should I create separate projects for each AI agent in my film workflow, or share one project across all agents?

Split by ROLE, not by agent. Start one project page for the creative producer agent that holds the script, shot breakdown, and characters — then spin a separate project page for every specialist role (storyboard, casting, costume, production design, DOP). One DOP per scene, two on a complex scene. Shared context lives upstream; per-role iteration lives in its own page.

invideo is an agentic video creation tool where you direct a crew of specialist agents inside project pages, and how you split those pages is the workflow decision that makes or breaks a multi-agent shoot.

Start with one creative producer agent in its own project page. This is the vision-holder. Load the full script, shot breakdown, and character details here once — every other agent inherits from this context. "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," says Hridaye, invideo's creative director. Skip this and your specialist agents drift because none of them share a source of truth.

Give each specialist role its own project page. Storyboard artist, casting, costume designer, production designer, DOP — each gets a separate page so feedback to one agent doesn't contaminate another's context. One documented 2-minute brand promo ran 8 specialist agents simultaneously across separate project pages and finished in 3 days for ~$1,500 (6,000–6,500 credits) — manual prompting equivalent would have taken at least a week, a traditional shoot ~2 months at $100K–$500K. Breaking agents across pages is what made that pace possible: targeted feedback per agent, no cross-contamination, parallel iteration instead of sequential.

Run multiple agents in the same role when the work demands it. Different scenes want different cinematography eyes — assign a separate DOP agent per scene rather than one DOP for the whole film. On a complex or long scene, put two DOP agents on the same sequence in parallel for two creative perspectives at once. Casting works the same way: a casting agent can run the same character prompt across two image models simultaneously on its page, and you pick the aesthetic you want before building character sheets.

Split the team across project pages too. A 3-person distributed team across multiple cities ran 3 separate projects in parallel inside invideo on a 5-day sprint — geography is irrelevant when everyone is working through the same agent surface, and separate pages give each person clean ownership. One reported personal setup used 6 agents simultaneously for a single short film.

Use these triggers to decide when to split a page off: (1) the role is different (storyboard ≠ DOP ≠ costume — always separate), (2) feedback for one agent would muddy another's context, (3) the scene is complex enough to want two agents on the same role, (4) a team member needs an isolated workspace to iterate without stepping on someone else's thread, (5) you're testing two models in parallel on the same creative question. If none of those apply, keep working in the page you're in — splitting for the sake of splitting fragments your context.

Across documented productions on invideo, multi-agent setups run 6–8 specialist agents in parallel across separate project pages, with one creative producer page upstream holding the script and character context. That structure — one shared vision page, many specialist pages — is the pattern that scales.

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.

— Hridaye, invideo's creative director