Digital compositing technique simulating depth in 2D footage — motion parallax and depth-of-field calculated in post. Cheaper than 3D reconstruction but obviously artificial at extreme angles.
You have flat 2D footage – an old shot, stock footage, or a low-resolution digital source – and suddenly need depth, parallax motion, or a noticeable depth-of-field effect. This is where the Noto process comes in. It reconstructs spatial information not through complex 3D modeling, but through intelligent pixel shifting and local deformation. The algorithm analyzes edges, textures, and motion patterns in the original frame and calculates how individual layers would move if the camera were actually moving through space. The result: synthetic parallax without real geometry.
On set, this saves you real production time. You shoot a static or minimally moving shot, and the compositor later creates the visual impression of an in-camera movement – a slight zoom-out imitation, lateral floating, or depth-of-field simulation. This costs a fraction of what a full 3D reconstruction would consume. For moderate effects (shallow focus, subtle camera parallax of a few percent), the footage looks quite clean. You only recognize the process at more extreme angles or when objects overlap each other – that's when it becomes artificial, because the algorithm has no real geometry and cannot solve occlusion "correctly".
Practically, it works like this: in the editing or compositing suite, you load the footage, roughly mark the "depth layers" (foreground, midground, background) – either manually or through automatic segmentation – and the process then calculates the displacement vectors. Modern implementations use machine learning to recognize edges more intelligently. This saves you considerable time compared to point tracking or manual rotoscoping. A good use case: old archival footage in modern productions where you subtly need more spatial dynamism without reshooting the scene.
The limit is where real object geometry becomes visible: if a person changes their arm position, the process cannot "understand" this – it only sees pixel displacement. Extreme camera zooms or rapid pans lead to artifacts at edges and in fine details. For a shot that lasts 5-10 seconds and requires at most 10-15° of "virtual camera movement", Noto is solid. Beyond that: it's better to work with real 3D reconstruction or even reshoot.