…a deepfake prank machine.
The model is trained to place a face into a styled portrait, nothing more. Guests see themselves reimagined, not impersonated.
You know the one: squeeze into a curtained box, snap four poses, wait five minutes, walk out with a strip that curls in your pocket. This is something else.
The model is trained to place a face into a styled portrait, nothing more. Guests see themselves reimagined, not impersonated.
No backdrop to hang, no keying to get wrong. A compact setup that fits in a corner and still looks like it belongs on stage.
The portrait appears on screen within seconds, and a QR code puts it on the guest's phone before they walk away.
Swap themes mid-event, schedule one look for the afternoon and another for the evening, or drop a surprise set when the dance floor fills.
Run it yourself or hand it to us. The setup is the same either way; only the amount of help changes.
LIGHT handles weak signal with queueing, caching and automatic retries. LOCAL sidesteps the problem entirely: it runs on one machine in the room and needs no connection at all.
Guests see a version of themselves they want to show someone. That reaction is the whole product.
Theme rotations keep the queue resetting: people come back for the version they have not tried yet.
Your branding is baked into every image, so it travels with each share.
Self-serve or fully managed. The system behaves identically; you choose how much of it you touch.
Here’s what stays true no matter the venue size, show flow, or chaos level.
Want the technical version instead? There is one further down.
Two very different machines behind the same thirty-second experience.
LIGHT runs as a horizontally scalable service behind a distributed job queue, so simultaneous capture and generation bursts across dozens of venues never collide. Traffic is load-balanced through a zero-downtime deployment pipeline, and an adaptive polling layer keeps each booth responsive even when LTE saturation or congested convention-centre Wi-Fi introduces heavy jitter. Automated janitors prune stale renders, sweep orphaned storage and rotate snapshots in the background; auto-rescue routines replay stalled jobs without the guest ever noticing.
LOCAL takes the opposite approach: there is no queue to stall and no server to reach. The face model runs directly on the neural engine of a single machine in the room, producing a portrait in roughly two tenths of a second. The installation brings its own network, boots unattended after a power cut, restarts its own services if one dies, and deletes every guest image automatically. Built to run for weeks without a keyboard attached.