To give the impression of moving in a 3D environment,
must create all the objects that surround us from
thousands of rectangular or triangular faces (verticles).
Ignatz with 3D Town had the idea to save it
to display a single image which rotates around the camera.
The human eye does not know all these faces
but derived objects that surrounds based on
color gradients.
Thus a person who approaches a pallet consists of
pink, brown, black, for example, dressed in jeans made
a palette of for example blue.
Depending on the location of the light source, blue is
lighter in front and behind and vice versa.
The idea is that the GPU guesses the object and its faces
and therefore its perspective and amends this according to the movement
of the camera. Calculation using shaders.
There should be minimum 2 images (front, rear)
or 4 images (north, south, east, west) and lateral
or 8 max pictures (8 directions on the numeric keypad)
to compose an object and not thousands.
Yes, if you want to animate people walking, for example, usually there are 4 or 8 sets of images, one for each direction. It is not as easy to do for things like buildings, though.
It’s difficult to explain my mind.
Imagine a 3d dice. Each face represents a man.
We can only see the man in 90° 180 270 360
Is it possible to display the man in other °
With extrapolation ? Like morphing between 2 pictures. the color dark or not can help us to realise perspective too. Thanks
Inferring inbetween angles from a set of flat images is likely to be prohibitive on performance and not very visually pleasing. If you want to render flat images, then old school games went for 8 directions of sprites. If you want it to look better/more modern, then I think the only way is with real 3d meshes.
If using 3d meshes, you need to worry about the quantity of vertices and faces, as this can have performance impact, but in the 3d world there are a number of techniques such as bump mapping that allow you to keep the vertex count on a mesh lower while giving good “high res” visuals.