Build a product mood board for AI advertising as a 9-image grid of hero shots covering the product at multiple angles, lighting setups, and compositions — then decompose that grid into structured prompt rules (style, materials, lighting, mood, exclusions) that every downstream image and clip generation inherits. It anchors brand visual language before any character or scene work begins.
Start by loading the project context. invideo is an agentic video creation tool that holds brand identity, visual language, and references in a persistent context tab — upload your website, color palette, tone of voice, product catalogue images (close-up, front, side, back, in-hand for scale), and any film or campaign references you want the look pulled from, and the invideo agent reuses all of it across every generation in that project.
Step 1 — Generate the 9-image hero grid. Ask the invideo agent to produce a mood board as a 3x3 grid of hero shots of the product across multiple angles, lighting setups, and compositions. This single output becomes the visual anchor for the campaign — done before any character, scene, or motion work. Nine images is the documented output size, and producing it as a grid lets you review the visual language side-by-side rather than judging shots in isolation.
Step 2 — Decompose each tile into prompt rules. Don't just keep the grid as pictures — convert it into language the downstream generations can inherit. For each tile, pull out one attribute: pose from one, lighting direction from another, depth structure from a third, color grade from a fourth, surface/material treatment from a fifth. Stated as conversational rules ("I want the depth structure from tile 3, the side-raked warm key from tile 6, the painted-backdrop falloff from tile 8"), the invideo agent holds each reference's specific nuance separately instead of blending them into mush.
Step 3 — Write the exclusions explicitly. A mood board for AI advertising is defined as much by what you reject as what you keep. List the failure modes the agent must never produce — plastic-looking surfaces, generic AI faces, harsh sunlight if the board is soft-light, white-on-white if the product is cream, any look that doesn't match the board. These standing don'ts live in the project context and stop the most common AI quality failures at the source.
Step 4 — Add a product sheet for scale and consistency. Alongside the aesthetic mood board, attach a product reference sheet: every angle of the product plus a hand-holding shot so the model learns true scale relative to a human hand. For intricate products (jewelry, layered packaging), validate consistency at three focal distances — close-up, mid, wide — before committing to a full generation run. As Hridaye, invideo's creative director, puts it: "Generate a close, a mid and a wide — with the product in every one. Confirm the piece holds before you commit the run."
Step 5 — Route mood board elements to the right model. invideo holds every current image and video model, and the invideo agent routes each task to the right one — you don't pick a platform per model. For mood board stills where lighting carries the look, Nano Banana renders lighting best; for tiles that need on-image text, typography, or design overlays, GPT-Image-2 is the choice. A practical two-model pattern for product hero stills: build the base image in GPT-Image-2 for aesthetic, then run Nano Banana to lock the exact product into it. For motion that has to inherit the board's language, Seedance 2.0 carries reference frames cleanly across shots; Kling 3.0 and Veo are alternates for specific motion needs. If one model's output starts feeling generically AI, render the same prompt across several models in parallel and pick the one that holds the board's language.
Step 6 — Probe before you scale. Generate one full-frame shot using the mood board rules — product, lighting, set, composition all in one frame. If that probe holds the look, the system is ready to batch the full campaign. If it drifts, fix the rules in context before committing credits.
For cost orientation: documented productions using this kind of mood-board-anchored workflow ran $33 for a full multi-platform brand campaign (155 credits), ~$150 for a 40-still / 30-clip editorial campaign (630 credits), and ~$2,400 for a three-ad jewelry campaign holding 100% product consistency across intricate pieces — the range reflects scope, not the method.
Watch some of these to see what works for you:
Generate a close, a mid and a wide - with the product in every one. Confirm the piece holds before you commit the run.
— Hridaye, invideo's creative director