Coopen / Work / AI ad generation
Our own platform
An AI that learns your brand and hands you the finished ad.
Most "AI marketing" tools give you a blank prompt box and call it a feature. Coopen Marketing does the opposite: Novah reads the business's own website, learns its brand, and produces the whole ad — the copy, the image and the branded template — ready to publish.
See it work
Thirty seconds, start to finish
The clip walks the actual product: Novah pulls the brand straight from a website, writes the ad, generates the image and drops it into the business's own template. What comes out the other side is a finished piece, not a draft to fix.
It is vertical because that is where the ads it makes end up — the format is the point, not an accident of filming.
Audio in English · 30s
The challenge
A small business does not have a marketing department. It has an owner who already works ten hours a day and is now also expected to feed a dozen social accounts with things that look designed.
Generic AI writers made that worse, not better. They produce text with no idea what the business sounds like, no logo, no palette, nothing that matches what the brand already looks like — so someone still has to open a design tool and rebuild it. The bottleneck was never the words. It was everything around them.
What we built
What we built
- Brand ingestion — Point it at a website and it extracts the identity: palette, tone, what the business actually sells. No forms to fill in before the first useful output.
- Novah, the assistant — Turns that identity into finished ads — copy and image together, composed on the brand's own template.
- Brand asset library — Logos, photos and templates stored per workspace, with AI tagging so the right photo gets picked for the right ad instead of hunted down by hand.
- Editor — Because the last 10% is always human: move, recolour, swap a photo, then export.
- Multi-workspace — An agency runs many brands without them ever leaking into each other.
Why "learn the brand first" is the whole trick
Any model can write ad copy. What decides whether the output is usable is everything the model is not told: that this business uses this blue, that its logo has clear space around it, that it never says "cheap" but does say "honest pricing".
Feed that in first and the generation stops being a lottery. Skip it and you have built a very expensive way to produce work someone else has to redo.
The stack
Built on Cloudflare's edge
Worker API, D1 for the data, R2 for the assets. Image generation and language models behind our own gateway, so keys never reach the browser.
The service behind it
Related
Still designing every post by hand?
If producing one on-brand ad takes your team an afternoon, that is automatable today. Show us your brand and we'll show you what comes out.