GENEYEAI LABS
1,000,000
SURFACE / 03 · DIGITAL TWIN

Headsupp.
Your digital twin,
always on.

WhatsApp for your digital twin. Create yours from one photo and about fifteen seconds of your voice — then friends, fans and clients can call it by text, voice or video, in any time zone, and it answers live, in your voice, with your face. And it remembers every person it talks to.

FILE / 03.A

What it does.

01

A twin from one photo.

Upload one photo and record about fifteen seconds of your voice. That single photograph becomes roughly 1,600 rendered frames, plus twenty expression clips — nods, smiles, listening looks — so it looks alive between words. No studio, no shoot, no reshoot. Setup takes about a minute.

02

It answers live, in your voice.

People reach your twin by text, voice or video, in any time zone. It listens, thinks and answers during the call, with your face and your voice. Built multilingual, so it can talk with people beyond your own language.

03

It remembers every caller.

Your twin picks up where each person left off, fan by fan. That memory stays between the two of them — one person’s conversation never reaches another.

04

You stay in control.

Every call shows an AI twin label, so people always know. You set its personality, what it knows and what’s off-limits. It can’t go public until your face matches, and you can pause, revoke or delete it at any time.

05

One link, shared anywhere.

Your twin gets its own link. Drop it in a bio, a story or a video and people tap straight through. Start private, test it yourself, and go public only when it feels like you.

06

Our own models, where it matters.

The voice and the face run on our own models, adapted from open weights for live conversation and hosted on our own GPUs. Quality, cost and your likeness stay in our hands.

FILE / 03.B

How it works.

A call runs a live loop: your twin listens, thinks, speaks and appears, then connects over WebRTC on a global edge. Speech-to-text and the conversation model are third-party and swappable; the two parts that define the experience are ours.

The voice and the face run on our own models, adapted from open weights for real-time conversation and hosted on NVIDIA A100 GPUs. Keeping those in-house is what keeps quality, cost and your likeness in our hands rather than a vendor’s.

Per-person memory is loaded at the start of every call, so the twin picks up where that person left off — and only that person. It is real-time and working end to end today, in private beta, with latency and concurrency the next things we’re tuning.

// On the term “digital twin”

The term digital twin was coined by NASA in the 1960s Apollo program — a “living model” that mirrors a real-world system with live data to monitor it, anticipate failures, and test fixes before they’re needed. Headsupp applies that same idea to people: a living, private model of you. Read NASA’s account →

You can’t be everywhere.
Your twin can.

Headsupp is in private beta and invites are going out now. Requesting one takes about two minutes — if you’re selected we’ll set your twin up with you, free while we’re in beta.

OTHER SURFACES

Also from GenEye AI Labs.