PAID1:56The anchor — the post this playbook is read from.
Open the post page →
InVideo creators upload project material, let an Agent establish the working context, and then direct a creative generation or revision.
Who runs it mattfarmerai, Joe Urbinati, and Gayatri Agrawal.
Not dissected yet. The captured frames and transcript are the available evidence.
These moves recur across the cited posts. Wording and execution vary; this is not one combined video timeline.
I uploaded a PDF of my story, a gladiator vs alien film.
Open the source post →
The script, character descriptions, reference images, just drop them all in.
Open the source post →
First go to ai.invideo.io, click on this agent mode and here is where you'll upload your script and treatment.
Open the source post →
The agent read the whole thing, pulled out my characters, my world, and my treatment, then asked where I wanted to start.
Open the source post →
NVIDIA Agent 2 reads it all, understands the context, and remembers it throughout the project.
Open the source post →
Agent one reads it, extracts the characters, world and treatment.
Open the source post →
Then I said I want my alien more venomous. Make your skin crawl. Insectoid alien horror. It redesigned the character, and updated every connected shot in my sequence.
Open the source post →
This playbook belongs to a structure shared across brands. Its member videos count toward each structure’s total. Open a structure to compare the grouped playbooks.
Now, after that, I gave one instruction. Start shot four, the night reveal.
Open the source post →
You can also then suggest edits like I want a human self and alien cover to feel like two different people. And wait look at this, it has suggested three approaches.
Open the source post →
What varies: Materials, agent roles, visual targets, and edit requests vary.
Scope: Exclude claims about memory, consistency, or filmmaking capability without the shared upload, context, and instruction sequence.