How the AI Works

This page explains how Guisera actually works: what runs where, what leaves your computer, and what the technology can and cannot do. We would rather you understood it than be impressed by it.

1. What the service is

Guisera is a real-time AI video transformation. Your camera feed is processed by a generative model that renders you as a chosen character while you are live. It is not a filter or an overlay laid on top of your face — the model generates each output frame.

That is why it needs your camera light on, why it costs money per minute, and why the output is never quite identical twice.

2. Where the processing happens

The transformation is performed by a third-party AI processing provider, not on your computer and not on our servers.

When you start a session:

1. Our server checks your balance and issues a short-lived access token for the provider. It expires within minutes and is limited to that session.
2. Your application connects directly to the provider and streams your camera video there. It does not pass through our servers.
3. Your chosen reference image — your Skin — is sent to the provider at the start of the session, as the target for the transformation.
4. Transformed video is returned to your application, shown to you, and passed to your virtual camera if you are using one.

Skins are stored on your device only. They are not uploaded to our servers and are not in our database. But they are transmitted to the processing provider each time you use one in a session. We say this explicitly because "stored on your device" is easily heard as "never goes anywhere", and that would not be true.

3. What we know about your sessions

What reaches our servers is the meter, not the content: that a session started, that it ended, how long it ran, how much of it produced no video, and what it cost. That is what your credit history is built from.

We do not record your camera feed or the output. We could not produce footage of your session if we were asked to, because none exists.

The AI tools that work on single images are different. Those images are sent to our servers and on to a provider for processing, and the result is stored so you can retrieve it from your history.

4. Your data is not used to train models

We do not use your video, your Skins, or your generated output to train or fine-tune AI models, and we do not sell them to anyone who would.

5. What the technology cannot do

Being straight about the limits saves you money and disappointment:

• It is not instant. There is a warm-up at the start of every session while the model loads. You are not charged for that time.
• There is latency. Video makes a round trip to a data centre and back. It is fast enough for conversation, not for anything requiring exact timing.
• Quality depends on your setup. Your camera, your computer, and above all your upload bandwidth. A poor connection degrades the result before anything else does.
• It is not deterministic. The same Skin and the same lighting will not produce a pixel-identical result twice.
• It can fail mid-session. A model can stall or a connection drop. Time that produced no video is not billed — see Credits and refunds.

6. Synthetic media, and telling people

The output is synthetic. It is generated, not filmed.

Many platforms now require you to disclose synthetic or manipulated media, and some jurisdictions are legislating for it. Meeting those obligations is your responsibility. Our conduct rules set out where we draw the line: presenting as a character is fine, making someone believe a specific real person is present is not.

7. Bias and representation

Generative models learn from data, and that data is not evenly distributed. Results may be less accurate for some faces, skin tones, lighting conditions, and camera angles than others. This is a known limitation of the technology and not something we can fully correct from our side.

If the service performs noticeably badly for you, please tell us at info@guisera.online. We would rather hear it than have you assume it is your camera.

8. Changes to the underlying technology

We may change the AI provider or model we use, to improve quality, reliability, or cost. If a change materially affects what happens to your data, we will update our privacy policy and tell you.