Static ads built from ads that already work.
Wisry's research agent reads the image ads running in your market on Meta and TikTok, extracts why they convert, and rebuilds them around your product, your palette and your claims. You approve a board of finished creatives; they publish straight into a Meta ad set. No brief, no designer queue, no blank canvas.
~200
static ads a month on Studio's 1,216 credits, if you spend the balance here
6
evidence-backed angles per research run, each carrying its cited sources
Meta + TikTok
ad libraries read end to end before a single concept is written
$0
of your ad spend taken as commission, while Studio is flat at $99 a month
How it works
Paste your store URL.
Setup starts from one link and nothing else. Wisry reads your catalog, your voice, your visual identity and your audience into brand memory, and every agent downstream works inside it. No ad account to connect, no performance history to upload, no cold start.
Agents read your market's ads.
A research agent pulls the static ads your competitors are running on Meta and TikTok and reads the layout, hook, offer and proof out of each one. Every finding sits next to the ad it produced.
Six angles come back, each with receipts.
The strategist turns that evidence into six campaign concepts, sorted by awareness stage. Each carries its hook, its message and links to the ads it came from, so you can argue with an angle before you pay to render it.
Concepts become finished statics.
Pick the angles worth running. The creative studio generates image ads for each one against your product facts and brand memory, in the sizes Meta serves, and lays them out on a QC board for review.
Approve once, launch to Meta.
Approved creatives group into ad sets and publish straight to Meta from the board. No export, no zip file, no Ads Manager. The optimization agents take it from there, moving budget toward whatever earns it.
Start from evidence, not a blank prompt.
Most AI ad tools begin by asking what you want to make. Wisry begins by finding out what is already selling. It searches the Meta and TikTok ad libraries for your category, assembles a pool of competitor creatives, and reads each one for the structure underneath it: the hook that stops the scroll, the offer doing the work, the proof carrying the claim. Video ads in that pool are transcribed and read the same way, so a hook that works in motion can be rebuilt as a still. Nothing is inferred from a category average; it is read off the specific ads your buyers are being shown this week. The blank page never appears, because the brief is assembled from ads that have already paid for their own testing. That is the difference between a generator and a research tool that happens to generate: one guesses at a market, the other reports on it before it asks you to spend anything.
- Search competitor brands by name, or let the agent find them from your category and your catalog
- Source ads collect into a reusable pool you can filter, pin and re-run future concepts against
- Every finding is anchored to the specific creative it came from, viewable full size on the board
- Research runs are budgeted up front, and the credits are refunded when a run returns no evidence
Six angles, sorted by who they're for.
A run returns six concepts, not one. Each is mapped to an awareness stage, so you can see at a glance which are prospecting to people who don't know the category exists and which are closing buyers already comparing two brands. That distinction is where most ad accounts leak money, and it is the thing a generic ad generator has no way to reason about. You edit the hook, tighten the message, swap the product, or send an angle back for another pass, all before any pixel is rendered. Two weak angles get discarded in thirty seconds and the four you believe in go forward, which makes this the cheapest place in the whole process to be wrong. It is also the review step most ad tools skip entirely, which is why so many of them feel fast and produce so little. Six is not an arbitrary number either: it is roughly the count of distinct creative hypotheses a small ad account can test in a fortnight without starving each one of data.
- Hook, message, offer and product arrive as editable text, not as a locked black box
- Awareness-stage mapping keeps prospecting and retargeting creative genuinely distinct from one another
- Refine an angle conversationally and regenerate only that concept, rather than the whole run
- Discard the two you don't believe in and pay to render only the four you do
Variants at ad-set scale, on a board you control.
Approving a concept produces a set of finished statics rather than a single hero image: enough variants to fill an ad set and give Meta's delivery system something real to choose between. They land on a QC board with the concept and its evidence beside them, so reviewing is a scan rather than a hunt through a downloads folder at the end of a long week. Product imagery comes out of your own catalog, so the bottle on the ad is the bottle you ship. An ad showing a colorway you don't sell is a returns problem, not a style choice. A retake costs a fraction of a full concept, which means raising your standard on one image is never an argument with your budget. Everything stays attached to the angle that produced it and stays in the workspace, rather than scattering across shared drives where nobody can tell version four from version seven.
- Regenerate a single variant without re-running the concept or repeating the research behind it
- Product imagery is pulled from your own catalog, not stock photography and not invented
- Accept, retake or reject each creative individually, with the source evidence still on screen
- Placement sizes match what Meta actually serves, so nothing needs a crop pass before launch
The last mile is the part nobody automates.
Generating an image is the easy half of the job. The expensive half is getting it live, keeping track of which hypothesis it belonged to, and reacting when the numbers arrive. Wisry publishes approved creatives into Meta ad sets from inside the product, with no export, no zip file and no re-uploading into Ads Manager at eleven at night, and keeps the concept-to-ad mapping intact, so every result maps back to an angle. Google campaigns run from the same workspace, on the same credits and the same brand memory. From there the optimization agents watch delivery, move budget toward the creatives earning it, and queue the next batch as the running set fatigues, a job that rewards checking every few hours and punishes checking once a week. You find out which of your six angles was right rather than which of twelve unlabeled JPEGs won, and that answer is what makes the next run better.
- Publish straight to Meta ad sets without leaving the board or opening Ads Manager
- Angle attribution survives the handoff, so every result reads as an answer to a question you asked
- Budget shifts toward the creatives that are working, continuously and without a meeting
- Fresh creative is queued from the same evidence pool as the running set starts to fatigue
Built for people who have to ship ads, not decks.
Brand memory, not a style prompt
Your palette, typography, product photography and tone of voice are read once from your store and reused by every agent from then on. You are not re-describing your brand in a text box for each new ad, and you are not correcting the same drift twice. Change something in brand memory and the next run inherits it, including runs an agent starts on its own.
Claims checked against your facts
Ad copy is generated against the product facts on file. Anything a source ad asserts that your product cannot support is flagged before the creative renders, not after it has spent money. The competitor's claim is evidence about how the market is being sold to. It is not permission to repeat it.
Every angle shows its sources
Each concept links back to the competitor ads and content it was built from, viewable full size. When an angle underperforms you can see precisely what it was betting on and retire that assumption instead of writing off the whole run and starting over from nothing.
One credit pool, spent how you like
Studio's 1,216 monthly credits are fungible. Put the whole balance into static ads in a launch month and into research or video the next, with nothing stranded in a per-feature allowance. Top-ups start at 25¢ a credit, get cheaper as the amount grows, last 365 days and survive cancellation, so an unusually heavy launch month doesn't strand credit you already paid for.
Your first board is one store URL away.
Paste your link, let the research run, and see six angles built from the static ads already winning in your category, with the sources attached.