Kive keeps campaign visuals consistent by storing the direction instead of the output: a studio holds the lighting, framing and environment, and a trained brand style holds the look, so the next product inherits both rather than being re-interpreted from a prompt. That distinction is the whole game. A campaign does not fall apart because one image is bad – it falls apart because forty images were each decided separately, and nothing carried the decision forward.
What a brand kit has to hold for AI work
A traditional brand kit is a colour list, a logo pack and a typeface. That is sufficient when a human designer sits between the guidelines and the artwork, because the human supplies the judgement the document leaves out.
Generated imagery removes that intermediary. The guidelines have to be executable, which means the kit needs to hold four things rather than two.
The four parts
First, the studio: lighting state, framing distance, environment. This is the part that decides whether image forty matches image one, and it is the part that never appears in a conventional brand document.
Second, the trained style. A model trained on your own images reuses a specific look across generations, which is a stronger guarantee than describing that look in words and hoping the next person words it the same way.
Third, the references the direction was approved from – the board someone said yes to. Fourth, the rule for which asset is canonical, so a later question about the approved version has an answer that does not depend on memory.
Where campaigns actually drift
Drift is not dramatic. Nobody generates a wildly wrong image and ships it. What happens is that the third product gets made on a Thursday by someone who was not in the approval meeting, and the backdrop comes out a different colour. When a person is in the frame the same thing happens to the face, which is a separate discipline covered in keeping the same model across AI product photos.
Three places it enters
The backdrop is the most common, because it occupies the most pixels and the eye compares it instantly across a grid. Shadow direction is second: two assets lit from opposite sides read as two shoots even when everything else matches. Crop distance is third and the most overlooked – the same product framed tighter looks like a different product line.
The image below is a real example from this article's own set. The perfume and the lipstick above share a backdrop and a lighting state. The sunglasses do not: same idea, different ground, and in a campaign grid it would be the frame that breaks the row.
What Canva and Midjourney store instead
Both tools hold something. Neither holds the thing that keeps a campaign together, and the gap is documented on their own pages rather than a matter of opinion.
Canva's Brand Kit is a container for brand assets – colours, logos, typefaces – applied to layouts. The free tier allows one Brand Kit, capped at three colours; Pro at US$180 a year – US$15 a month on yearly billing – allows five Brand Kits; Business at US$250 a year per person allows a hundred Brand Kits and adds approvals [1]. That governs the design layer wrapped around a picture. It has no opinion about the light falling on the product inside the picture, which is where a campaign visibly breaks.
Midjourney gets closer. A style reference carries the visual vibe of an image – colours, medium, textures, lighting – and its documentation is explicit that it does not copy objects or people [2]. Style weight is tunable from 0 to 1000, defaulting to 100. So the look travels and the product does not, which for product campaigns is the wrong half.
A style code is not a saved direction
The reuse mechanism matters as much as its quality. Midjourney's style codes stick across a reroll or a variation, but its docs note that under permutation, repeat, or Draft mode each image comes back with a different style code [2]. Style reference versions are also incompatible with Moodboards, so the board a direction was approved on cannot be the thing that enforces it.
That is the difference between a parameter and a stored decision. A parameter has to be carried by whoever runs the next generation, correctly, every time, in a compatible mode. A saved studio is selected rather than retyped, which takes the operator out of the consistency equation.
None of this makes either tool a bad choice. Canva is the right answer for laying out the ad once the imagery exists, and Midjourney is the right answer when the goal is an aesthetic rather than a specific product rendered accurately forty times.
- 1Kive
- Best for
- Producing a campaign set where the product must stay accurate
- Pricing
- Basic $20/month, 1,000 credits, about 100 generations
- A studio stores lighting, framing and environment and is reapplied, not retyped
- Trained brand styles hold the look across products
- Approved references and the generated set live in one library
- Not a layout tool – captions, logos and CTA still happen elsewhere
- Heavier setup than typing a prompt for a single image
- 2Canva
- Best for
- Laying out and resizing the ad once imagery exists
- Pricing
- Free (1 Brand Kit, 3 colours); Pro US$15/month billed yearly, US$180/year (5 Brand Kits); Business US$20.83/month per person billed yearly, US$250/year (100 Brand Kits)
- Brand Kit enforces colours, logos and type across layouts
- Approvals and multi-team management on the higher tiers
- Brand Kit governs the design layer, not the lighting inside the image
- The free tier's kit is capped at three colours
- 3Midjourney
- Best for
- Chasing an aesthetic rather than a fixed product
- Style references carry colour, medium, texture and lighting
- Style codes and the Style Explorer make a look discoverable
- Documented not to copy objects or people, so your product is not preserved
- Style codes change per image under permutation, repeat and Draft mode
Building the kit in the order that works
The sequence matters more than the tooling. Deciding the look after producing twenty assets means redoing twenty assets.
From an approved board to a repeatable kit
Approve the direction on a board first
Collect references and settle the argument before any production. Our guide on making a mood board covers the collecting half; the point here is that the board is the input to the kit.
Translate the board into one studio choice
Name the lighting state explicitly – even flat, soft overhead, hard sun, low key – and pick the preset that holds it. This is the single decision the whole campaign inherits.
Train the style if the look is specific
Where a brand look cannot be described in a sentence, Kive documents training a custom model from your own images so it can be reapplied rather than re-explained.
Generate the whole set against that kit
Run products through the same studio rather than deciding per asset. Consistency then happens by construction instead of by discipline.
Review the set side by side, not one by one
Lay the frames out at the same size. Backdrop tone, then shadow direction, then crop distance – in that order, because that is the order the eye notices them.
Cut the placement variants from the approved set
A feed square, a story vertical and a landscape crop are three framings of one decision, not three new jobs. Derive them from assets already generated against the kit so the backdrop tone is identical across placements.
Kive's documented brand controls cover the pieces this sequence needs: reusable and custom visual presets, trained brand styles, prompt libraries and references, with collaborative review around them. Basic is $20 per month for 1,000 credits – roughly 100 generations – and unused credits roll over up to twice the monthly limit (kive.ai/pricing).
The last step is where kits usually leak. Someone recrops the hero for Reels in a different tool a week later, the export lands on a slightly different white, and the paid placement stops matching the product page it clicks through to. Keep the crop inside the same set the kit produced, and treat a new aspect ratio as a framing of an approved asset rather than a reason to regenerate. The mechanics of reshaping an approved frame without regenerating it are in AI image extender and layer editing for products, and platform-side requirements for those placements are in our guide to AI product videos for TikTok and Meta ads.
The check that catches drift before launch
One test, run before anything ships: put every asset in the campaign on one screen at the same size, in the order a customer will meet them. Not a folder of thumbnails – the actual sequence, actual scale.
Drift is invisible asset by asset and obvious in a row. If one frame pulls your eye for the wrong reason, it is almost always backdrop tone or shadow direction rather than anything about the product. Our guide to brand colour accuracy covers the colour half of that check, and AI studios covers how the lighting half gets stored. For the approval stage that feeds the kit, see how to make a mood board.
Consistency has a reputation as the boring requirement, the thing that constrains the work. In practice it is the thing that makes a small brand look like it has a creative department.
A customer never sees your brand kit. They see six images in a row and form an opinion about how carefully the company is run, in less time than it takes to read this sentence. Storing the direction is how you keep that opinion from depending on who happened to generate the third asset.
“Kive seems to read my thoughts and knows what suits the URBAN BAG brand best.”
Studios that hold a direction
Live from Kive Discover. Store the look once so the third product cannot drift.
References
