AI campaign assets are the finished files a campaign ships: the ad with the price on it, the flyer, the line sheet, the social set in four sizes. A model returns a picture. Turning that picture into an asset takes layout, type, brand marks, a legal line and an export pass. That last step is a canvas job, and it is the one most teams still do in a second tool.
You have 40 generated product shots and a launch on Thursday. Every one of them is beautiful and none of them is shippable, because not one carries the price, the logo, the offer or the size the placement wants.
So the files go to a layout tool. Then a second version comes back for approval, then a third, and by Friday nobody is sure which PNG is the live one. This guide covers what a generated image is missing, where that work belongs, and how to stop the round trip from eating the time the generation just saved.
Key Takeaways
- Generation ends at the picture. Price, logo, call to action, legal line and export sizes are all still ahead of you.
- The round trip is the real cost. Exporting to a layout tool and back is where brand drifts and version sprawl starts.
- Type belongs on the canvas, never in the prompt. Text you lay out is editable and sharp. Text a model draws is neither.
- One canvas, two destinations. Screen exports and print exports come off the same file at different resolutions.
- Multi-page is the forgotten deliverable. Line sheets, one-pagers and sales decks are campaign assets too, and they need PDF, not PNG.
- Budget the disclosure line. Where a rule or a platform asks for an AI label, that label is part of the asset.
What counts as a campaign asset
A campaign asset is a file someone can upload without opening anything else. That definition is stricter than it sounds, and it is why a folder of generations is not a campaign.
Run the test on any file you are about to hand over:
| Element | Comes from the model | Comes from the canvas |
|---|---|---|
| Product image | Yes | No |
| Background, scene, lighting | Yes | Adjustable after |
| Headline and body copy | No | Yes |
| Price, offer, promo code | No | Yes |
| Logo and brand marks | No | Yes |
| Call to action button or badge | No | Yes |
| Legal or disclosure line | No | Yes |
| Placement sizes | Partly | Yes |
| Print-ready resolution | Partly | Yes |
Five of the nine rows are pure layout. That is the work sitting between a good generation and a live ad, and no prompt closes it.
The five things a generated image is missing
Type you can edit. Ask a model for a price on a pack shot and you get pixels shaped like a price. Change the promotion and you regenerate the whole image. Type placed on a canvas is a text object: change 20% to 30%, re-export, done.
Hierarchy. An ad works because the eye lands in an order. Headline, product, offer, action. That order is a layout decision made with position, scale and contrast, none of which a text prompt controls reliably.
Brand marks at the right size. Your logo is a file with a defined clear space and minimum size. It gets placed, not described.
The size set. Meta’s published Feed image guidance puts the Feed at 4:5, 1440 x 1800 pixels (facebook.com/business/ads-guide, as of September 2026), while Stories and Reels are full-screen vertical and in-stream is landscape. One generation does not cover that. One canvas, exported four ways, does.
A print path. Trade shows, packaging inserts, retail POS and wholesale line sheets all need resolution measured against physical size, not screen pixels, and a PDF rather than a PNG.
What the round trip between tools costs
Every hop between generation and layout adds three things: a copy of the file, a chance for the brand to drift, and a person who has to be asked which version is current.
That compounds quietly. A ten-asset campaign becomes 40 files across two tools and a shared drive. Nobody deleted anything, because nobody was certain. The six-tool AI stack runs the same arithmetic across a whole toolchain.
Keeping the layout next to the generation removes the hop. The DesignerBox image editor puts images, shapes and text on one canvas, with your Assets in the sidebar. You relight a shot, change an angle or remove a background inside the editor, without exporting anything (designerbox.ai/image-editor, September 2026). It makes one edit after another, and the picture keeps its detail and resolution.
Background removal there has an in-browser version and a server-side AI version for difficult edges (designerbox.ai/image-editor, September 2026). The AI version uses credits, and the cost is shown before the run. Hair, glassware and fabric fringe are the usual reasons to use it. The six-photo test for AI background removers checks those edges before you choose a tool.
What belongs in the prompt and what belongs on the canvas
The rule is simple: if it might change next week, it goes on the canvas.
Prompt it: the product, the surface, the lighting, the scene, the camera angle, the mood. These are photographic decisions and the model is good at them, and a moodboard generator is where a team agrees them before anyone spends a credit.
Lay it out: every word, the logo, the price, the badge, the button, the disclosure, the frame, the crop for each placement.
Teams get this backwards and pay for it in regenerations. A campaign with five price points does not need five generations. It needs one generation and five text layers.
Templates shorten the layout half further. The image editor ships presets in categories such as Social, Slides, Marketing, Flyers and Posters, with multi-page designs carrying per-page content (designerbox.ai/image-editor, September 2026). Starting from a template and swapping the generation is faster than building the frame each time, and it is the same logic that makes AI photo templates worth locking on the generation side.
Export: one canvas, screen and print
Two destinations, two settings, one file.
Screen. PNG when you need transparency, JPG when you do not. Sized per placement. If you are handing to a scheduler or an ad platform, name the files by placement before you export, not after.
Print. PDF, and resolution chosen against the physical size rather than the pixel count. The image editor exports multi-page PDF at 72, 150 or 300 DPI (designerbox.ai/image-editor, September 2026). Use 300 for anything that gets printed and handled, 150 for large-format viewed at distance, 72 for screen-only PDFs like a deck.
Multi-page PDF is the export most teams forget they need. A wholesale line sheet, a retailer one-pager, a trade-show leaflet and a sales deck are all campaign assets, and all of them are several pages of the same brand system. Building them as separate PNGs and assembling elsewhere puts the seam right back.
For the channel-by-channel version of the resize step, the ads resizer makes one ad in every size the ad platforms need. Check the focal point on each size, because a plain centre crop can cut the product.
The disclosure line nobody budgets for
Where a law or a platform asks for an AI label, the label sits inside the asset, which means it sits inside the layout. A disclosure is type, in a defined position, at a legible size, on every variant that needs it.
The rules differ by market and by what the image shows, and they change. Not every AI image needs a label. This is general information, not legal advice. The practical planning point is only this: decide the wording once, put it in the template, and it appears on every export automatically. Retrofitting a line across 40 finished files is the version of this that hurts. The current state of the requirements, and which images trigger them, is covered in labelling AI-generated images.
How to run it
- Generate the picture only. Product, scene, light. No text, no logo, no badge.
- Open one canvas per campaign, not per asset. Multi-page keeps the set together.
- Place the fixed furniture first. Logo, safe margins, disclosure position. That is your template.
- Drop the generation in and set the hierarchy. Headline, product, offer, action.
- Duplicate the page per placement and adjust the crop and the type scale to each shape.
- Export twice. Screen sizes as PNG or JPG, anything printed as PDF at 300 DPI.
- Version by changing text, never by regenerating. New price, new page, same picture.
Steps 1 and 7 are the ones that save the budget. Everything else is craft.
For the surrounding picture, AI digital marketing across all six functions sets out which parts of the operation AI produces and which it only assists.
Where this fits in the campaign
If the ad itself is the deliverable rather than a layout around it, a product ad template builds the composed creative from a product photo, and the canvas becomes the place you refine it. For the social and open-graph end of the set, an app for link preview images covers the share cards that otherwise get made last and worst.
The volume question, how many finished assets a campaign needs each month, is worked through in producing 100 ad variants without a team. The category-level view of what these tools do and do not automate sits in what an AI ad generator is.
Build the layout once and the rest of the catalogue is a text change. The template that carried the first product carries the next forty. The frame you set on Thursday runs again on every SKU after it, and the type sits in the same place on every product. That is what turns finishing one ad into finishing a season of them.
In DesignerBox, the result and the layout stay in one workspace, so there is no folder between two tools. Start from a template, add your brand and your products, and run it. The cost is shown before the run. See the templates.
FAQ
What is the difference between a generated image and a campaign asset?
A generated image is a picture. A campaign asset is a file someone can upload without opening anything else, which means it also carries the copy, the logo, the offer, any required disclosure and the right dimensions for its placement. The gap between the two is layout work.
Should I ask the AI model to put text on the image?
No. Text drawn by a model is pixels, so changing a price means regenerating the image, and the type rarely holds up at small sizes. Place text on a canvas instead, where it stays editable and sharp and where you can version an offer in seconds.
How do I make one design work across every ad placement?
Build one master layout, then duplicate it per placement and adjust the crop and type scale. Meta’s Feed guidance is 4:5 at 1440 x 1800 (Meta Ads Guide, as of September 2026), Stories and Reels are full-screen vertical, in-stream is landscape. Exporting from one canvas keeps the artwork identical across all of them.
What resolution do I need for printed campaign assets?
Choose the resolution against the finished physical size rather than the pixel count. The DesignerBox image editor exports multi-page PDF at 72, 150 or 300 DPI (designerbox.ai/image-editor, September 2026). Use 300 DPI for anything printed and handled, 150 for large-format seen at distance, 72 for screen-only PDFs.
Can I edit a generated image after placing it on the canvas?
Yes. The DesignerBox image editor runs background removal, relighting, angle changes and skin retouching inside the editor, so you do not export a shot to fix it and import it again (designerbox.ai/image-editor, September 2026).
Does background removal cost credits?
The server-side AI version uses credits, and the cost is shown before the run. It is the one worth using on hair, glass and fabric edges (designerbox.ai/image-editor, September 2026). The credits in each plan are on the pricing page.
How do I build a line sheet or a multi-page PDF?
Use a multi-page design with per-page content and export the whole thing as PDF (designerbox.ai/image-editor, September 2026). Line sheets, retailer one-pagers and decks are campaign assets like any ad, and they need the same brand furniture placed once and repeated per page.
Sources
- DesignerBox image editor page, canvas features, templates, in-editor AI edits and export formats (designerbox.ai/image-editor), accessed September 2026
- DesignerBox ads resizer page (designerbox.ai/apps/ads-resizer), accessed August 2026
- Meta Ads Guide, Facebook Feed image aspect ratio and resolution (facebook.com/business/ads-guide), accessed September 2026
- DesignerBox pricing page (designerbox.ai/pricing), September 2026
Product capabilities from DesignerBox product pages, and placement specifications from Meta’s advertising guide, as of September 2026. Disclosure requirements differ by market and change; verify against current rules for your market before publishing. Individual results vary.
