Real time, photorealistic, driven by your own face
LookingPro sits between your face and your meeting. It reads your face frame by frame, renders a photorealistic version of you driven by that motion, and delivers the result as a standard camera device. Nothing changes inside the meeting software.
Capture, not generation
Audio-driven systems infer mouth movement from speech. They produce a talking image and lose everything that was not spoken: the pause before an answer, the raised eyebrow, the moment of hesitation that tells a client you are thinking rather than reading.
LookingPro is driven by the face itself. The camera tracks expression, gaze and head motion, and the rendered image follows that motion. Lip-sync comes from the way your mouth actually moves, not from a model's estimate of how a mouth should look while a word is spoken.
Real time is the hard part
Rendering a convincing face offline is a solved problem. Rendering one during a conversation is not. Capture, tracking, rendering and encoding all have to fit inside a frame budget while people are talking, and any instability shows up at once as a glitch or a visible seam.
Our current builds run at 720p and 1080p, and we test them against the video conferencing software people actually use, Zoom and Google Meet among them. Holding that budget, while keeping the image steady and free of seams, is the part of the system we invest in most heavily.
Delivered as a camera
LookingPro registers as a camera device on Windows and macOS. Any application that can select a webcam can select LookingPro.
No plug-in, no new platform
There is no plug-in inside the meeting software, no change to the way your organisation runs calls, and no dependency on a single vendor's platform.
Where the work happens
The rendering runs on our servers. The application sends your camera input over an encrypted connection, receives the rendered image back, and presents it to your meeting software as a camera. This is what allows the same quality on a laptop that does not carry a workstation graphics card.
LooperPro, for the short interruption
Not every interruption needs a rendered proxy. LooperPro captures you at the moment you switch it on and plays that back as a loop, using the same virtual camera path, so the meeting software treats it as an ordinary camera you selected.
What we do with that data, and for how long, is set out in the terms you accept in the application before enrolment.
Two disciplines that rarely meet
Photorealistic faces come from visual effects practice. Knowing that a face belongs to the person claiming it comes from biometrics. LookingPro is built by people from both: founders who built and sold a biometric and deepfake detection company, working with facial animation specialists whose credits come from feature film production.
That combination shapes the engineering. Rendering a face is common. Rendering one that runs inside a live conversation, looks photorealistic and stays tied to the person it belongs to is a different problem, and it is the one we work on.
Questions about the engineering
We are glad to answer technical questions: how the virtual camera behaves inside your meeting software, what hardware the pipeline expects, and how the two products differ in practice.
Write to tech@lookingpro.com, or use our contact page.