Alpshade
AI-based Visual Anonymisation

Seen, not identified.

We replace individuals with photorealistic artificial characters while preserving the scene.

Same scene, subject anonymised

Drag the slider to compare the original footage with the Alpshade version.

OriginalAlpshade

On the right, the original subject has been replaced with an artificial character that preserves the attributes of the original, such as ethnicity, gender, expressions and body language. Everything else in the scene stays as filmed.

Based in Switzerland

Swiss studio, Swiss technicians and Swiss machines for processing your data.

Proprietary workflow

Developed in-house by Alpshade and cleared through a major public broadcaster's security review.

Local first

Undisguised footage stays on local machines. It is never uploaded to a cloud AI service.

Never used for training

Your footage is never used to prompt or train an AI model.

The problem

Blur was made for the human eye. Software sees through it.

Blurring, pixelation and silhouettes were designed to stop a viewer recognising a face. Today, freely available AI tools can sharpen a blurred face back into a recognisable one.

11/15

common anonymisation methods were reversed by AI in a 2022 academic study. Gaussian blur was among the most exposed.

The only reliable protection is to take the real face out of the footage. That is what Alpshade does.

Sources: arXiv 2210.10651 · arXiv 2506.12344.

Work

A real person on screen. Nobody you can find.

Our proprietary workflow keeps the framing, light and body language of the original shot, so the scene still plays as testimony.

BBC World Service from 2:58

Broadcast work

Shadowmen

A BBC World Service documentary about Russian operatives who say they are sabotaging Putin's war. The anonymised contributors appear from 2:58.

BBC World ServiceFull subject replacement

Full case footage from broadcast investigations is shared in a briefing, under NDA.

Security

The contributor's safety is the brief.

Alpshade's proprietary workflow was built around one question: How do we protect people visually in the AI era?

Jurisdiction

Data is processed on Alpshade machines in Switzerland by Swiss-based technicians, under Swiss data-protection law.

Undisguised footage

Stays on local machines. It is never uploaded to a cloud AI service.

Cloud tools

Only masked or pre-anonymised material ever reaches a cloud tool. Those sessions are deleted after sign-off.

Model training

Your footage is never used to prompt or train an AI model.

Face search

Every generated face is checked with independent face search so it doesn't resemble a real person. You receive a PDF report.

Who we work with

For anyone who needs to show people without exposing them.

Documentary productions

Protect contributors, keep the story

Investigative and current-affairs teams keep at-risk contributors on screen without revealing who they are. Their testimony, expression and presence stay intact.

Marketing

Customer testimonials, kept private

Feature real customer stories when the customer would rather not show their face. Their words and delivery stay authentic, and their identity stays private.

Tech companies

Personal data out of your visuals

Replace identifiable people in images and video before you store, share or use them, so your visual data keeps working without exposing anyone.

FAQ

Questions producers ask us.

How is this different from blurring or a silhouette?

Blur and pixelation leave the real face in the footage, just degraded, and AI tools can often recover it. Our proprietary workflow removes the real person from the frames entirely and puts a newly generated character in their place. There is nothing left to reverse.

Will the stand-in look like my contributor?

The character fits the scene: similar age, build and styling, so the shot stays believable. The face and identifying features are new, and we run face search to confirm the character doesn't match a real person.

Can you go through our supplier security review and sign our NDA?

Yes. Our workflow has already passed a broadcaster's information-security review, and we will complete yours. We are happy to share our data-handling documentation before any footage changes hands.

What do you need from us to start?

The clips, a short brief about the contributor and the risk they face, and your delivery spec. We reply with a recommended approach and an estimate for your shot list.

How long does it take?

It depends on the number of shots and the level of protection needed. After a briefing we give you a schedule per shot list, and we can run several projects in parallel when a deadline is close.

Should we tell the audience the person was replaced?

We recommend a clear on-screen disclosure. We can suggest wording that fits your broadcaster's editorial guidelines.

Contact

Tell us about your next investigation.

Conversations are confidential. Send no footage until we have agreed how it will be handled.

Email
Keep it general. We'll agree a secure channel before you share anything sensitive.