Have a close look at the first two photographs. Those are four strip light portraits composed into one family portrait
Although this is quite some popular editing style, here is a common mistake. All individual photographs are taken with the same distance to the models. When composing the models appear to stand next to each other but partly behind each other so that the distance to the camera appears different in the composing. By doing so the propotional size of he models does not seem to match reality.
This effect is not related to displaying he smallest head to the left or in front of the others respectively. To demonstrat that I composed the faces from the smallest to the largest and vice versa.
Why is that?
See the following simplified illustration.
All shapes are composed next to each other with the same size, but our brains expect those to be smaller that should appear farer away than the closer ones.
When you think of it during the photo session then you could position the camea in various distances to match the perspective for the later composition.
However, I did not think at that from start.
If you repeat my mistake and take all photographs from the same distance. or if you don't know the composing sequence from start, then you need to scale them accordingly. In my example the individual images were about 5000px in height. I scaled those to the left (appearing mor to the back) by 300px each resulting in heights of 4700px, 4400px and 4100px.
You may well argue that this level of scaling may not be perfectly accurate but you will agree that the result looks much more realistic.
When I do a similar shoot next time then I'll go for he option of taking the shots with different distances from start. In the end it saves a lot of work and uncertainies.