Every shot generated before a brand’s product and look are actually locked is a shot that risks drifting from every other one. A commercial isn’t one clip, it’s a sequence, and a campaign built on top of it needs cutdowns for half a dozen platforms and markets that all still need to look like they came from the same shoot.

Locking isn’t a formality before the real work starts, it is the real work. What a model is given before generation begins, real product photos, a brand book, a reference spot, decides how consistent every shot after it will be, far more than any prompt written mid-generation can fix.

Here’s what actually goes into locking a product and a look before a single shot gets generated.

Why locking comes before generation, not after

Generating a first draft and fixing inconsistencies afterward almost always costs more time than locking things upfront. Once a product’s proportions or a brand’s color grade have drifted across a few shots, correcting them means regenerating, not adjusting, and each fix risks introducing a new inconsistency somewhere else.

Invideo Agent treats the brief, the brand book, and the product references as a locked foundation before any shot generates, rather than something to refine after seeing a first draft. The reference spot that started the conversation gets broken down for structure and pacing before it becomes a template for the new ad, not copied blindly.

What actually goes into a locked product setup

A product that shifts in size, color, or material between shots is the most common way a commercial’s consistency falls apart. Locking it starts with using real product photos, packaging shots from every angle the film will need to show, rather than pulling an image from a website. A hand-holding shot gives a model an accurate sense of scale, and describing how the material actually behaves, soft and matte versus hard and reflective, gives it something concrete instead of a guess.

Invideo Agent locks the first shot of a product setup before generating anything else, then runs a dedicated pass specifically for holding that product steady, separate from the pass generating the surrounding scene’s aesthetic.

What Agent Two adds: reading a reference spot and rebuilding it

The newer invideo Agent Two model can read an existing reference ad directly and break it down into its actual components, the structure, the pacing, the turn, how the product lands in frame, rather than treating the reference as a vague mood board.

It also reads reference uploads more precisely when each one arrives with a specific note attached. A reference tagged “this one’s for the lighting only” gets used only for that, instead of the model pulling every visual trait from it indiscriminately, which is the difference between a handful of well-annotated references and forty unlabeled ones.

Locking the brand’s look, not just the product

A brand’s look is more than its logo and color palette. Splitting it into two references keeps it from getting muddled: a director’s bible covering the grammar, camera language, lighting, color, tone, and a separate production bible tracking the operational details, shot list, lens choices, location locks.

A brand book itself can be handed over directly rather than translated into a written prompt. Feeding in an existing document, a brand guideline, a treatment note, lets the system extract the actual rules, colors, tone, likes and dislikes, instead of a director having to describe them from scratch in text.

Common problems when the lock isn’t tight enough

Product drift is the most common failure, usually because the reference was a low-quality or website image rather than a real photo with a clear sense of scale and material.

Brand tone slipping between shots is the second most common issue. If only the product is locked and the brand’s color grade or camera language isn’t documented anywhere, later shots tend to drift toward a generic look instead of the brand’s specific one.

Partial locking, sheeting a product’s front angle but not its back, or locking color but not lighting, creates inconsistency in exactly the attributes that were left undocumented. A lock is only as complete as what was actually included in it.

How this fits into a full AI ad films workflow

Locking the product and brand is the second of six steps in producing a full commercial, after the brief comes in and before casting, storyboarding, and generation begin. Skipping ahead to generation before this step is complete tends to show up later as inconsistency that’s harder to fix than it would have been to prevent.

The same locked setup carries through to the AI ad films pipeline’s later steps, so a product locked at the start stays consistent through casting, every board setup, and the final cutdowns built for different platforms.

Common mistakes when locking a product and brand

• Using website product images instead of real photos. Website images rarely give a model an accurate sense of scale or material behavior.
• Skipping a hand-holding reference shot. Without one, a model has to guess a product’s true size relative to a person.
• Uploading many unlabeled references at once. A reference with no note attached risks pulling every visual trait from it rather than the one quality intended.
• Locking the product but not the brand’s camera language or color grade. A commercial can hold a product steady and still drift in tone if the brand’s look was never documented.
• Treating generation as the step where consistency gets fixed. Correcting drift after the fact usually takes longer than locking references properly before the first shot.

FAQ

What counts as a “locked” product setup? At minimum, real product photos from every angle the film will need, a hand-holding shot for scale, and a physical description of how the material behaves. The first shot of a setup should also be locked before any variant or follow-up shot generates from it.

Do I need a formal brand book to lock a brand’s look? An existing brand guideline or treatment document helps, since it can be handed over directly rather than rewritten as a prompt. Without one, a director’s bible covering camera language, lighting, color, and tone serves the same purpose.

How is locking different from just writing a detailed prompt? A prompt describes what should happen in one generation. Locking establishes reference material, real photos, a broken-down reference spot, a brand book, that stays available and consistent across every shot in the project, not just the one it was written for.

What happens if only part of the product or brand gets locked? Whatever wasn’t documented tends to drift first. Locking a product’s front angle but not its back, or its color but not its material behavior, leaves exactly those gaps open for a model to guess at differently across shots.