White-Label Content AI Platform

Quick answer — A properly white-labelled platform runs on your domain, sends from your mail server, stores files in your bucket, and calls AI providers with your keys — so your customers never see the vendor and your data never leaves accounts you control.

Vitra.ai Universe supports this at organization level; white label covers the surface.

Three levels, and only one is real white-labelling

Level one is a logo. Your mark in the corner of someone else's product. The URL still says the vendor, the emails still come from the vendor.

Level two is branding. Logo, product name, primary colour, support email, and the application served on your own domain with its own certificate.

Level three is infrastructure. Your storage, your mail server, your AI provider accounts. At this level the vendor is a piece of software you run rather than a service you route customer data through.

Most platforms offer the first two and call it white-label.

What is configurable at organization level

Provider preference is the underrated one

Bring-your-own-key usually means one key for one vendor. Preference ordering is different: you choose which provider handles each job — text to speech, speech to text, voice cloning, audio separation — and in what order to fall back.

That matters commercially. If you have negotiated rates with one voice vendor, route voice there. If a provider is unavailable in a region you operate in, route around it. The platform does not decide for you.

Translation memory can be yours too

Memory is where the accumulated value sits — every approved translation your team has ever signed off. Being able to keep that on a provider you control, rather than inside the vendor, is what makes leaving possible.

Translation memory supports VitraTM alongside Phrase and XTM, chosen per memory rather than per organization — so an agency already standardised on XTM keeps it, and a brand with no TMS gets one.

The question to ask any vendor

Not "can I put my logo on it." Ask where content is processed, whose account the model call is billed to, and whether you can export the memory and the glossary in a standard format.

One caveat worth knowing about storage: once files exist in a bucket, the credentials tend to lock, because moving storage under a live system means moving every asset. Decide the bucket before you load it, not after.

Where to start

Configure the domain and the branding first, because they are visible and quick. Then work through the infrastructure list in the order of what your security review will ask about, which is usually storage, then keys, then mail.

FAQ

What does white-labelling an AI platform actually involve? Three levels. A logo, then full branding on your own domain with your own mail server, then infrastructure - your storage bucket and your own AI provider keys. Only the third means data stays in accounts you control.

Can you use your own AI provider accounts? Yes, at organization level. You supply keys for the model, speech and transcription providers you use, and set a preference order deciding which vendor handles each job and what it falls back to.

Can translation memory be kept on your own provider? Yes, chosen per memory rather than per organization. That matters because the memory holds every translation your team has approved, and controlling it is what makes switching platforms possible.

What should you ask a vendor about white-labelling? Not whether you can add a logo. Ask where content is processed, whose account the model call bills to, and whether the translation memory and glossary export in a standard format.