TAKE Performance and growth teams

Creative Automation Software: Dynamic Creative Optimization and DCO Tools for Video Ads

Two completely different products are sold under the phrase creative automation, and dynamic creative optimization is a third thing that is not creative production at all. One resizes a creative you already own. One writes a creative you do not have yet. DCO permutes whatever pool you hand the ad platform. Most teams buy the first, expect the second, and blame the third when volume stays flat. This page separates all three, prices them, and shows the asset math that decides which one you are short of. Or skip ahead: paste a product page URL on the right.

Published rate cards, marked as such Platform asset caps from the docs Variant-count claims, discounted

Ad Recipe

TAKE

Pick a Creator

Hook Style

Free to start - no credit card required

The short answer Last updated August 2026

Creative automation is software that produces ad creative without a designer touching every file, and it splits into two categories that share a name and solve opposite problems. Versioning tools take one approved master and mass-produce sizes, languages, offers and feed-driven variants: Celtra, Bannerflow, Storyteq, The Brief (formerly Creatopy), Adobe GenStudio. Generation tools produce creative that did not exist, including new hooks, presenters and scripts. Dynamic creative optimization is neither. DCO lives inside Meta and Google, and at delivery time it assembles combinations from the asset pool you uploaded, then shifts spend toward the winners. That distinction decides your budget, because DCO is a multiplier on your asset pool, not a source of assets. Hand Meta three videos and five headlines and the reporting will show dozens of combinations, but there are still three videos, and in a video-first feed the video is what carries the result. The bottleneck is almost always upstream of the tool people buy to fix it. Teams that solve it by commissioning more shoots run into a different wall, which is what video production company pricing does to a per-variant budget.

Total asset cap, Meta asset feed

Max videos in a Meta asset feed

6.67

New creatives a week at $50k spend

$49

UGCGen, per month flat

Read this first

Creative automation is two products sold under one phrase

The category grew out of display advertising, where the real pain was mechanical. A brand approved one banner concept and then needed it in forty sizes, nine languages and six offers, refreshed whenever the price changed. Software that could take a master file, a template and a product feed and emit two thousand correct files was genuinely transformative, and that is what the established creative automation platforms were built to do.

Then paid social moved to video and the pain moved with it. What a growth team runs out of now is not sizes of one idea. It is ideas. Meta rewards a steady supply of genuinely different hooks, presenters and openings, and no amount of resizing produces a second hook. A versioning platform handed a single concept returns that concept in more shapes, which is exactly what it promised and not at all what the buyer needed.

Both products are legitimate. The failure is in the shopping, not the software. Before you compare vendors, decide which sentence describes you: I have the creative and I need it everywhere, or I have the channels and I have nothing new to put in them. Almost every disappointed creative automation deployment we hear about is a team that answered the second and bought for the first.

	Versioning automation	Generation automation
What it produces	One approved idea in many sizes, languages and offers	Ideas that did not exist: new hooks, presenters, scripts
What it needs from you	A master creative, a template and usually a product feed	A product page, a photo or a rough brief
Typical vendors	Celtra, Bannerflow, Storyteq, The Brief, Adobe GenStudio	AI ad and UGC generators, including UGCGen
Where it shines	Large catalogs, many markets, strict brand governance	Hook testing, creative fatigue, teams with no studio
Where it fails quietly	You ship 400 files that are all the same ad	Brand consistency needs a human pass before launch
Buying process	Mostly sales-gated, annual contracts, onboarding project	Mostly self-serve, monthly, live the same afternoon
Honest limitation	Cannot invent a concept, and does not claim to	Will not run your global brand system across 40 sizes

Vendor placement reflects each product's primary design intent as described in its own marketing. Several vendors now sell into both columns, and the generative features bolted onto versioning platforms are improving quickly.

The expensive misunderstanding

Dynamic creative optimization is a multiplier on your asset pool, not a source of assets

DCO is the reason a lot of teams believe their creative problem is solved when it is not. It is a delivery-time system inside the ad platform. You upload a pool of components, the platform assembles combinations per impression, measures them, and pushes budget toward what works. On Meta this shows up as dynamic creative and as the Advantage+ creative enhancements; on Google it is the responsive formats and the asset groups that feed Performance Max and Demand Gen.

It works. Handing the machine the combination decision is almost always better than guessing, and the reporting is honest about what won. The trap is arithmetic. When the interface tells you it is testing dozens of creatives, it is counting combinations. Three videos crossed with five headlines and five body texts is seventy-five combinations and still three videos. In a feed where the thumb stops or does not stop in the first second, the video is the variable that decides the outcome, and the count that matters is three.

This is why creative volume is a supply problem rather than an optimization problem, and why buying more optimization does not fix it. The platform will happily optimize a thin pool forever. It has no mechanism to tell you the pool is thin, because from inside the auction a small pool with one decent performer looks like a working campaign right up until that performer fatigues and there is nothing behind it.

	Creative automation	Dynamic creative optimization
Where it runs	In your stack, before the campaign is built	Inside Meta or Google, at delivery time
What it changes	How many assets exist	Which of the existing assets each person sees
Can it originate footage?	Yes, that is the entire job	No. It recombines and transforms your uploads
Who you pay	A software vendor, monthly or annually	Nobody. It is a free feature of the ad account
What limits it	Budget, brand review, production capacity	The platform asset caps and the pool you supplied
Fixes creative fatigue?	Yes, if it generates rather than versions	No. It reallocates spend inside a fatiguing pool
Honest summary	Decides what exists	Decides what gets shown

DCO is a free platform capability, so treating it as a purchase is itself a category error. The paid DCO vendors sell orchestration, feed logic and cross-channel reporting on top of it, not the optimization itself.

The number that settles the argument

The platforms publish a hard ceiling on how much creative they will accept

If DCO were really a substitute for creative supply, the ceiling would be high or absent. It is neither, and the number is public. Meta's marketing API documentation for the asset feed states a maximum of 30 total assets, and within that, no more than 10 images, 10 videos, 5 bodies, 5 titles, 5 descriptions, 5 call to action types and 5 links. The per-type limits add up to 45, so the 30 asset ceiling binds first and you cannot max out every type at once.

Read the video row again. Ten. That is the entire creative variety a single Meta dynamic creative ad set can hold, forever, no matter how large your budget is or how good the optimizer gets. An account spending 300,000 dollars a month and an account spending 3,000 both get ten video slots. Optimization is not the scarce resource here and the platform is telling you so in its own developer docs.

Google draws the line in the same place. Performance Max takes a minimum of one video and a maximum of 15, with 3 to 15 headlines and 2 to 5 descriptions. Responsive display takes 1 to 5 videos per orientation. Demand Gen takes 1 to 5 headlines, at 40 characters rather than the 30 most guides claim. These are not generous allowances waiting to be filled. For most accounts they are a target that has never once been hit.

Surface	Video slots	Text slots	Overall cap
Meta dynamic creative (asset feed)	Up to 10 videos, and up to 10 images	5 bodies, 5 titles, 5 descriptions, 5 CTAs	30 assets total, which binds before the per-type caps
Google Performance Max	Minimum 1, maximum 15, each 10 seconds or longer	3 to 15 headlines, 1 to 5 long headlines, 2 to 5 descriptions	Per asset group, with image counts listed per aspect ratio
Google responsive display	1 to 5 videos per orientation	1 to 5 headlines, 1 long headline, 1 to 5 descriptions	1 to 15 images horizontal and 1 to 15 square
Google Demand Gen	Documented recommendation of 3 per orientation, no stated maximum	1 to 5 headlines at 40 characters, 1 to 5 descriptions	1 to 20 images per orientation, carousel 2 to 10 cards

Meta figures are from the marketing API asset feed specification. Google figures are from the Google Ads Help asset requirement pages for each campaign type, checked August 2026. Google lists image counts per aspect ratio and does not publish an aggregate, so do not multiply them. Platform limits change without much announcement, so re-check before you build against them.

One nuance worth stating precisely, because a lot of writing on this topic gets it wrong in our favor and we would rather not benefit from that. Meta's Advantage+ generative features do produce new media. Image animation turns a still into a short video, video uncrop extends footage into another ratio, and text optimizations rewrite copy you supplied. Every documented feature is derivative of an asset you uploaded, and nothing in the documentation originates footage from nothing. So the accurate claim is not that the platform creates no media. It is that the platform permutes and transforms a pool you supply, and the pool is capped at ten videos.

The case against everything above

The best evidence on creative fatigue argues against constant refresh

We sell creative volume, so treat what follows as a company arguing against its own pitch, and weigh it accordingly. Nearly every number you have read about creative fatigue is unsourced. The frequency threshold of 2.5 or 3.0, the claim that conversion likelihood falls 45 percent after four exposures, the rule that you need one fresh asset every one to two weeks: we went looking for primary sources for these and did not find any. They trace back to vendor blogs citing other vendor blogs. Meta's own creative fatigue help article cannot be read by an automated fetch, and Google publishes no refresh cadence at all. Its only hard numbers point the other way, asking for at least 14 days of ramp-up before an asset is judged.

The strongest independent research we could find actively disagrees with the refresh narrative. Kantar analyzed between 1,292 and 1,706 tracked ads across three markets and reported that regardless of how many times an ad is seen, "Fed Up" scores do not tend to increase. Wear-out, on that evidence, is far less automatic than the industry assumes, and a strong ad may have a much longer useful life than a fortnightly refresh schedule allows.

So here is the honest version of the argument. Volume is not valuable because ads decay on a timer. Volume is valuable because you do not know in advance which idea works, and the only way to find out is to put several genuinely different ones into the auction and let the platform tell you. That is a search problem, not a decay problem. It also implies something a creative vendor is not supposed to say: once you have found a winner, the correct move is often to leave it running much longer than a refresh calendar would suggest, and spend your production on finding the next winner rather than on replacing the current one.

Monthly ad spend	New creatives per week, median	Top quartile
Under $10k	2.80	4.83
$10k to $50k	4.10	8.09
$50k to $200k	6.67	15.95
$200k to $1M	11.24	31.11
Over $1M	18.85	54.64

Source quality: vendor-published. These figures come from Motion, which sells creative analytics and therefore has an interest in higher numbers. We include them because they are the only spend-banded creative volume benchmarks we could find with a named source, not because we think they are neutral. Treat them as a rough sense of peer behavior, not a target.

Discount this number, including ours

When a platform advertises a thousand variants, it is counting combinations

Variant counts are the headline metric of this entire category and they are close to meaningless as stated. A tool that produces four background colors, five headlines, five call-to-action labels and ten crop sizes can legitimately print the number one thousand on a slide. What it produced is one advertisement with a thousand serial numbers.

The number worth tracking is different and much smaller: how many genuinely distinct openings you can put into the auction this month. A distinct opening is a different first line, a different person saying it, or a different reason to care. Two ads with the same first three seconds and different button colors are one test, not two, and the platform will figure that out faster than the deck did.

Apply that filter to your own numbers before you apply it to a vendor's. Most teams who believe they shipped sixty creatives last quarter shipped six, versioned ten ways. That is not a criticism of the team, it is the predictable output of a stack built for versioning. It is also why the honest measure of a creative pipeline is distinct hooks per month, and why we report our own output that way rather than in renders.

Counts as one

Same footage, same script, resized to 9:16, 1:1 and 16:9, with three button colors.

Counts as one

Same opening line and presenter, with the price and the offer swapped from a feed.

Counts as three

Three presenters opening on three different objections: cost, fit, and does it actually work.

What it actually costs

Creative automation pricing, with the source of every number marked

The most useful thing about this table is the fourth column. Roughly half of this category publishes nothing at all, and the SEO pages that appear to know their prices are guessing. We checked each vendor on its own site in August 2026 rather than trusting a search summary, and where a vendor publishes nothing we say so instead of inventing a range.

The pattern is clean once you see it. Enterprise versioning platforms are sales-gated with annual contracts and an onboarding project. Self-serve generation tools publish rate cards in the tens of dollars a month and you can be live the same afternoon. That gap is not really about price. It is about whether the product assumes you already have a creative operation to plug into.

Vendor	What it is for	Published pricing, August 2026	Source quality	Worth knowing
The Brief (was Creatopy)	Versioning and templates	Create $29/mo, Ultra $79/mo, Team $49/seat/mo, all billed annually	Published rate card	Rebranded from Creatopy. Ultra includes 100k ad impressions a month
AdCreative.ai	Static and generative ad creative	Starter $39 to $189, Professional $249 to $400, Ultimate $999 to $1,399	Published rate card	Priced by credits and brand count, not seats. Quarterly 25% off, yearly 50%
Pencil	Generative ad creative	Core $14/mo for 50 generations, Growth $55/mo for 250	Published rate card	Cheapest published entry point in the category. Pro is quoted
Marpipe	Catalog and feed management	Feed Management free, Startup $199/mo, Enterprise from $999/mo	Published rate card	Now positioned around feeds and SKU caps rather than creative testing
Celtra	Enterprise versioning at scale	Nothing published	Sales-gated	Priced on platform, scale, impressions or video starts and add-ons
Bannerflow	Enterprise versioning	Grow, Pro and Enterprise tiers named, no prices	Sales-gated	Tailored estimate on request. Their /pricing path 404s, use /pricing-0
Storyteq	Enterprise versioning and DAM	Price on application	Sales-gated	One-off implementation charge plus an annual license, tiered by users
Hunch	DCO and creative automation	Nothing published	Sales-gated	Describes itself as pay for what you use with no seat limits
Smartly.io	Cross-channel DCO and media	Nothing published, and no pricing page exists	Sales-gated	Third-party estimates conflict badly. We are not repeating any of them
Adobe GenStudio	Enterprise content supply chain	Not verified	Could not confirm	Adobe hosts would not respond to us. Get this one from Adobe directly
UGCGen	Generative UGC video	Starter $49/mo, Plus $149/mo, Pro $499/mo	Published rate card	That is us. Counted in finished ads, free tier is watermarked

Checked on each vendor's own pricing page on 29 August 2026. Two corrections worth passing on: search results confidently reported an AdCreative.ai Ultimate tier at $599 for 40 users, and the vendor's own page shows Ultimate starting at $999 and priced by credits and brands rather than users. Creatopy no longer exists under that name. Pricing in this category moves, so confirm before you budget.

The pipeline that actually works

How to build a creative pipeline that feeds DCO instead of starving it

The order matters more than the tooling. Teams that get this right treat generation as the supply line and DCO as the distribution layer, and they never let the second run ahead of the first.

Write the objections, not the ads

List the four or five reasons a qualified buyer does not purchase: price, fit, trust, timing, whether it actually works. Each one is a hook family. This is the only step that needs a human who knows the customer, and it takes an afternoon.

Generate a distinct opening per objection

Produce one video per hook family with a different presenter and a different first line, from a product URL or a script. Five real openings beats fifty versions of one, and it is a smaller job than it sounds.

Load the pool, then let DCO decide

Upload every distinct opening into one ad set with your text variants and turn dynamic creative on. Do not split them across ad sets to protect a favorite. The machine is better at this than you are, provided you gave it something to choose between.

Replace the winner before it dies

Retire on trajectory rather than on collapse. When a hook family stops improving, generate the next two openings inside it and load them, so the pool is never rebuilt from zero under pressure.

The risk nobody budgets for

You cannot review a combination the machine has not built yet

Here is the part of dynamic creative that legal teams find late. Every ad DCO serves is an ad your company published, but no human approved most of them, because they did not exist at approval time. You reviewed a pool. The platform shipped the cross product of that pool, and it will keep shipping new members of it as long as the campaign runs.

For most brands that is harmless. It stops being harmless the moment the pool mixes an AI-generated presenter with first-person language, because the machine has no idea it just manufactured a testimonial. Pair a synthetic person with a text asset that reads "I have used this every day for six months" and the resulting ad claims a consumer experience that never happened, attributed to a consumer who does not exist. Nobody wrote that ad. You are still responsible for it.

The rule in play is the FTC trade regulation rule on consumer reviews and testimonials at 16 CFR Part 465, which took effect in October 2024 and is still in force. Section 465.2 prohibits, among other things, misrepresenting that a review or testimonial is by someone who does not exist or who did not have the experience described. It carries civil penalties per violation and it is being enforced: the FTC sent warning letters to ten companies in December 2025, just over a year after the rule took full effect. The rule is not about AI, which is exactly why it catches this. It does not care whether a machine or a person assembled the claim.

The practical control follows directly from the mechanics. You cannot approve seventy-five combinations, so you have to constrain the pool until no bad combination is constructible.

What is in your DCO pool	What the machine can assemble	Safer construction
Synthetic presenter plus first-person experience copy	A testimonial from a customer who does not exist	Keep experience claims out of the text pool entirely
Synthetic presenter plus product demonstration copy	A branded advertisement, which is ordinary	This is the default to build toward
Real customer footage plus generic brand copy	A genuine endorsement, with disclosure obligations	Keep the material connection disclosure in the asset
Star ratings or review counts as swappable text assets	A rating attached to a product it was never about	Bind ratings to the product asset, never to the pool
Results claims swapped independently of the product	A substantiated claim moved onto an unsubstantiated SKU	Pin claim assets to the SKU that carries the evidence
Anything implying the presenter is a paying customer	Exactly what 465.2 addresses	Label the presenter as a spokesperson in the creative

This is a plain reading of a published rule for planning purposes and it is not legal advice. If your pool mixes synthetic presenters with experience claims, have counsel look at the pool rather than at individual ads, because the pool is what determines the set of ads that can exist.

Read the constraint the other way and it is a buying argument rather than a warning. The rule closes the cheap shortcut, which is manufacturing consumer voices at scale, and leaves the legitimate method completely open, which is producing branded demonstration at scale. High volume was what the format demanded anyway. The compliant path and the effective path are the same path.

Who this is for

Where a generation layer pays for itself, and where it does not

In-house growth teams

Two or three people running seven figures of annual spend with no studio. The constraint is never the media buying, it is that the creative queue is one freelancer deep. Generation moves that queue from weeks to the same afternoon.

Performance agencies

Creative is what agencies get fired over, and it is the line item clients resist paying for. Producing hook volume in-house without a shoot day is the difference between a retainer that renews and one that gets reviewed. See white label UGC ads for the delivery side.

Large catalogs

Hundreds of SKUs and a handful of hero videos, so ninety percent of the catalog runs on static images. Per-product video at catalog scale is a volume problem, which is what bulk UGC content is built for.

Structured creative testing

Teams running a real testing calendar need a steady supply of variables worth testing. Feeding a matrix by hand stalls within a month. Our notes on ad creative testing cover the structure.

Seasonal and promotional cycles

Black Friday, back to school, a launch window. The creative is disposable by design and a produced shoot cannot be justified for a two week flight, which is exactly the economics generation was built for.

Where it does not fit

A brand film, a founder story, a category-defining campaign, or anything where the craft is the message. Hire people. Generation is the wrong instrument and no honest vendor should tell you otherwise.

Where the alternatives genuinely win

We generate the creative. We do not serve it.

We should be blunt, because the tables above could be read as a competitive claim against the versioning platforms and they are not one. We are not a creative automation platform in the established sense and we are not a DCO vendor. There is no template engine here, no product feed connector, no brand governance workflow, no approval routing, no digital asset manager, no cross-channel delivery and no per-impression assembly. If you need forty banner sizes in nine languages driven off a feed with locked brand rules, Celtra, Bannerflow, Storyteq and Adobe GenStudio built exactly that and did it well for years before we existed.

Nor do we compete with the platform DCO you already have. Meta and Google give it away, their optimization is better than any third party layer sitting on top, and the correct move is to use it. Everything on this page assumes you will.

What we do is narrow. We produce distinct video creative from a product page URL, a photo or a script, presented by AI creators with voiceover and burned-in captions, exported in 9:16, 1:1 and 16:9 at 1080p with commercial rights, at 49 dollars a month rather than a shoot day. That is the supply line into everything else described here. It is one layer of a stack, and a team with a working studio and a full creative calendar does not need it. A team whose DCO pool has been the same four videos since spring almost certainly does.

One more caution that applies to us as much as anyone. Treat every efficiency statistic in this category with suspicion, including the ones in vendor case studies that sound precise. They are published by companies selling the tool, the control group is rarely described, and the pilot is usually run by the vendor's own team on an account chosen because it was likely to work. Run your own two week test against your own baseline and believe that number instead.

Creative automation and DCO questions, answered

What is creative automation?

Creative automation is software that produces ad creative without a designer building every file by hand. It covers two different jobs. Versioning tools take one approved master and mass-produce sizes, languages and feed-driven offers. Generation tools produce creative that did not previously exist, including new hooks, presenters and scripts. Both are sold under the same phrase.

What is dynamic creative optimization?

Dynamic creative optimization is a delivery-time system inside an ad platform. You upload a pool of components such as videos, headlines and body texts, and the platform assembles combinations for each impression, measures which perform, and shifts budget toward the winners. It is free on Meta and Google, and it optimizes the pool you supplied rather than adding to it.

What is the difference between creative automation and dynamic creative optimization?

Creative automation decides what exists. DCO decides what gets shown. Automation runs in your stack before the campaign is built and increases the number of assets available. DCO runs inside Meta or Google at delivery time and chooses which of those assets each person sees. Buying more DCO cannot fix a shortage of assets, because DCO has no ability to create one.

What does DCO stand for in marketing?

DCO stands for dynamic creative optimization. In practice it names any system that assembles an ad from interchangeable parts at serve time and optimizes the combination against performance. The term covers both the free native features in Meta and Google and the paid third-party platforms that add feed logic, orchestration and cross-channel reporting on top.

How does dynamic creative optimization work?

You supply component assets into one ad set: several videos or images, several headlines, several primary texts. The platform generates combinations, serves them, and learns which pairings work for which audience segments, concentrating delivery on the strongest. The optimization is genuine and usually beats manual guessing. The ceiling is set entirely by how varied the pool you uploaded actually was.

Does dynamic creative optimization actually work?

Yes, at what it does. Letting the platform pick combinations reliably beats picking them yourself, and the reporting on which components won is useful. What it does not do is create variety. If your pool is three videos that all open the same way, DCO will optimize among three videos that all open the same way, and results will track the fatigue of that pool rather than the quality of the optimizer.

Does Meta have dynamic creative optimization?

Yes. Meta offers dynamic creative at the ad set level, where you upload media and text options and the system combines them per person, alongside the Advantage+ creative enhancements. Both are free with the ad account. Both are capped: the marketing API asset feed accepts a maximum of 30 total assets, of which no more than 10 can be videos and 10 images, with 5 each for bodies, titles, descriptions and calls to action.

Can AI generate ad creative automatically?

Yes, and this is the part of the category that changed most recently. Generation tools now turn a product page URL, a photo or a script into finished video with a presenter, voiceover and captions, in minutes and without a shoot. What AI does not yet do well is decide which idea is worth testing. That judgement is still the highest-value thing a human on the team contributes.

How many ad creatives should I test per month?

Count distinct openings, not renders. For most ecommerce accounts a workable rhythm is four to six genuinely different hooks a month, each tied to a different buying objection, with the weakest retired as they fatigue. Ten versions of one idea is one test. Teams that report shipping sixty creatives a quarter have usually shipped six, versioned ten ways.

Is dynamic creative worth it for small budgets?

It is free, so the cost question does not really apply, but small budgets do change the calculus. DCO needs enough delivery to separate combinations, and a pool that is too wide on a small budget spreads learning too thin to conclude anything. Under modest spend, run fewer components with real differences between them rather than many with cosmetic ones.

Does DCO create new ads?

It creates new combinations, not new footage, and that distinction is the most expensive misunderstanding in the category. Three videos crossed with five headlines and five texts reports as seventy-five creatives and remains three videos. Meta's Advantage+ features do generate some media, but every documented one is derivative of an asset you uploaded: animating your still, extending your video, rewriting your copy.

What are the best creative automation tools?

It depends which half you need. For versioning at enterprise scale with brand governance and feed integration, Celtra, Bannerflow, Storyteq and Adobe GenStudio are the established choices and all are sales-gated. For self-serve generation you can price yourself, The Brief, AdCreative.ai, Pencil and UGCGen publish rate cards and start the same day. Plenty of teams end up running one from each column rather than choosing.

How much does creative automation software cost?

The two halves price completely differently. Enterprise versioning platforms are sales-gated with annual contracts, so Celtra, Bannerflow, Storyteq, Hunch and Smartly publish no number at all. Self-serve tools do: as of August 2026 Pencil starts at 14 dollars a month, The Brief at 29, AdCreative.ai at 39, UGCGen at 49 and Marpipe at 199. Confirm with the vendor before you budget.

Give the optimizer something to optimize

Paste a product page URL or your own script, pick a presenter, and get a UGC-style video with voiceover and burned-in captions in 9:16, 1:1 or 16:9. Enough distinct openings to keep a dynamic creative pool honest, without a shoot day. Free to start with your account, no credit card needed.

Try the generator free See pricing

Keep reading

The testing side