HYGH built a camera-to-billboard preview tool on OpenAI's APIs

The interface that closes HYGH's core gap

HYGH connects more than 4,000 digital displays across Germany, from shop window screens to what it describes as the country's largest 3D LED billboard. Its stated business problem is visual: advertisers need to see what a campaign will look like on those screens before they commit, and producing that material used to be slow.

The company's answer is a campaign preview tool built on the OpenAI APIs. Per the source, an advertiser can snap a photo of a product and instantly see it displayed on HYGH's public screens. That is a specific frontend: a capture step, a generation step, and a rendering of the result in the context of a real screen placement.

It was a way longer process before to come up with material for pitches. Now we can spark client interest with custom creative work much faster.Montana Labs

This matters because the interface is the product for HYGH's in-house agency. The value isn't a chatbot window — it's collapsing the distance between a client's idea and a viewable mock-up on the medium HYGH actually sells.

Two paths from ChatGPT to shipped software

The announcement describes two distinct build tracks. The first is internal tooling: developers use Codex to bootstrap code, set up project files, and get prototypes running quickly. HYGH claims it launched five smaller internal tools in one week and now ships about two usable MVPs per week, versus the one-to-two months it cited before.

The second track is the outward-facing preview tool built directly on the OpenAI APIs rather than inside ChatGPT. That distinction is worth noting for anyone reading this as a frontend case study: ChatGPT Business accelerated the internal development loop, but the customer-facing experience required committing to API integration and owning the interface.

The creative workflow sits between the two. HYGH's team drafts copy and generates visuals in ChatGPT Business, then refines them in design tools or Sora. The output is human-finished; the model handles the blank-page problem that Link repeatedly names as the bottleneck.

Adoption structure, not just adoption

HYGH's account is unusually specific about the operating model. Employees already used ChatGPT personally, so the move to ChatGPT Business was framed as adding structure: a shared workspace, admin controls, and GDPR-conscious data handling — a real constraint for a European company.

The company runs weekly 'workflow Wednesdays,' where employees demo automations and share practices. Link's observation that people use AI at different scales, and that younger staff 'don't use it as Google, they live with it,' points to a deliberate attempt to spread uneven internal expertise rather than assume it.

Shared project links get a specific mention: teammates see not just the final output but how it was produced. For a company shipping two MVPs a week, that visibility into method is how a small team keeps reusable knowledge from getting lost.

What HYGH's preview tool signals for API-first frontends

The reusable lesson here is narrow and concrete. HYGH used ChatGPT Business to speed up how it plans and builds software internally, but its differentiated customer experience — photograph a product, see it on a public screen — lives in a purpose-built interface on the OpenAI APIs, not in a general-purpose assistant.

That split is the practical takeaway for teams building on these models: internal velocity and a shipped product frontend are separate problems. The first can be bought off the shelf and adopted bottom-up. The second still requires deciding what interaction you own, wiring it to your own data and screen network, and taking responsibility for what the user sees. HYGH's 'Internet of Public Screens' pitch only advances because it built that second thing.