Can you upload a style guide to an AI agent and have it apply your visual rules automatically across every shot?

Yes. Upload a complete visual-language document to the invideo agent once at project start, and it applies those rules — camera, lighting, palette, composition, mood — to every shot without re-prompting. One documented production loaded a 25-page director style guide as a permanent instruction set, and the invideo agent enforced a 9-element prompt structure across every frame of the film.

Load your style guide before you generate a single frame, and instruct the invideo agent to treat it as the rulebook for the whole project. invideo is an agentic video creation tool, and the invideo agent keeps an uploaded document in persistent context — it reads your treatment once and holds every directive across every shot and scene, checking each generated frame against the document before returning it. In one documented production, a 25-page style guide encoding Wong Kar-wai's visual system was uploaded as a permanent instruction set, and the invideo agent output 12 parameters per shot — film reference, shot design, lens, lighting plan, color script, atmosphere layers, blocking, final prompt, negative prompt, and more — all derived from the document.

Structure the guide so the invideo agent can act on it, not just store it. The documented version ran 14 sections: camera, angles, colour tone, atmosphere, mood, lighting, composition, movement, film palettes, prompt templates, negative prompts, and a quick-reference card. It also fixed a 9-element prompt assembly order — camera spec, lens and aspect ratio, lighting source, palette, composition, atmosphere, mood register, film/DP attribution, negative prompt — that held across every frame. Two details do disproportionate work: name your colour rules as tonal modes with exact hex values so palettes are reproducible, and include a "what never to do" section per segment of the film, which makes the invideo agent's autonomous decisions far more reliable.

The evidence that this is real rule application — not surface pattern-matching — comes from the invideo agent acting on the document unprompted. In one production the invideo agent applied a slow-shutter motion smear effect from page 17 of the document without being asked. In another, it flagged shadows leaning blue-green instead of the neutral gray the document specified, citing the relevant rule before the director noticed the deviation. It also pulled a named principle from page 12 and applied it to a scene type the document never specifically addressed. Once the document is loaded, a three-word continuation prompt — "Everything should match" — is enough for the invideo agent to carry character, lighting, lens grammar, and spatial logic across a multi-shot sequence.

Your style guide doesn't have to be a written document. One 2-person team uploaded 64 frames from a reference animation aesthetic in a single message with the instruction "I want you to deeply understand this art style and save it into context for further generations," then prepended the resulting style block to every prompt across 164 generated clips. Make negative constraints explicit in the block — that production's read "not live action, not photorealistic" — because prohibitions are what stop style drift mid-project.

Validate the guide before committing generation credits. Ask the invideo agent to apply the style to a genre or subject the director never worked in: if it asks clarifying questions and returns stylistically coherent output, the document has been internalized as grammar rather than aesthetics. Also challenge its technical claims — in one session the invideo agent had noted "anamorphic" for a director who shoots spherical at 2.40:1 hard matte, and corrected itself when questioned, before the error could propagate across the asset pipeline.

This workflow has held across documented productions of real scale: a 70-second film produced in 2 days for $750 with no style re-prompting, a ~90-second film built from roughly 400 video generations against a single director-protocol document, and a project with scenes numbered past 21 running on one loaded context. Note that the style guide governs visual rules; character identity is a separate layer, handled with character reference sheets the invideo agent holds in the same context — no LoRA fine-tuning required.

Watch some of these to see what works for you:

Agent One reads your treatment doc once and keeps it loaded across every frame. The thread stays held, scene to scene. No re-explaining. No starting over.

— invideo's creative team