Can't bring loose peg back in place. (was: "Making shadows stick to their sources?")

Another problem I’m having with quadmap-based shadows I’ve made for my scene (besides this one) is that they seem to be “detached” from their respective sources. This is more or less what an object and its shadow look like, when viewed from the top:

When I rotate the object, the shadow rotates in a completely different way. Also, when I make a character go up a ramp, the shadow separates from its feet and goes down…

I haven’t found a way to make the objects and their respective shadows share the same master peg… Is there any way of attaching these to the bottom of the objects they’re supposed to be coming from?

Any help would be really appreciated.

So, in the end, I’ve “solved” the detached character’s shadows by adding a peg to it and manually making it go up and down with the character, since I don’t know of an official way to make it stick to its feet when going up and down.

Now, regarding the objects’ shadows, I’ve brought them together (more or less) with the Translate tool on the Top view, but I can’t do the same with the pivot, which seems to be further away and causes a mess when I try to rotate the object and its shadow. Here’s an image that illustrates where it is:
(The object is in pink, in the Top view)

The closest thing I’ve managed is to move it with the Translate tool in the Perspective view, but that way it only moves temporarily and resets the next time I select the object.

I can only move it left and right in the Top view for some reason… could this be a bug, perhaps?

Any suggestions will be welcome

As I mentioned before, the shadow already looks fine with the right amount of blur, so that’s not an issue:

The problem happens when I want the characters to follow the path in red and the shadow keeps gradually separating from its “source”. =\

Were there pictures that were posted?

It seems like you are talking about a sample but nothing is visible in your post.

Oh, the example is actually those parallel lines you see in the post:

That’s exactly how an object and its shadow look like, from above (in the top view)…

Just throwing ideas out here… You have to address shadows separately for every camera angle. If this were a 3D program you would have the freedom to move a camera and see light behaving close to the way it does in reality. But a quadmap is just like you are taking a flat screen and laying it out so that it looks like it is displayed on a surface. However because it is not actually light blocked by an object with mass and form it is going to betray the shortcomings if the subject moves into certain positions.

If you need to see the same scene from the top, the top view has to be built from scratch because drawings are like sheets of paper. This means you need to draw as though you are looking from the top. then you have to use your judgement when positioning the quadmap so it matches the scene. The samples we see in videos are taking the most stereotypical shadow applications observed in our world. A quadmap can pull that off a lot of the time. There will be instances where you may have to use subterfuge to disguise the flaws.

The easiest situation for a quadmap is if you are looking straight on perpendicular to the object’s plane. It gets increasingly problematic the further you deviate from this.

EDIT: just to clarify, if you need a top view you draw your scene out as though it were seen from the top but you do it in the regular window not a top view camera. Top view cameras are useful for manipulating objects like moving props around and seeing where props are. But it is like being in the rafters of a theater stage with backdrop panels. You don’t actually use them to get a top view that is from the top of your animated world.

I’ve just made a couple of screenshots of the two variants of the problem I’m getting:

The first case is the object that shows a parallel shadow in the top view, as seen here:
If I move it in the Z axis, the parallel lines change their distance even. Really odd.

The second case is more simple. I want to make this character go up a ramp, but the shadow acts as if he is floating:

At least the second case might be a really simple thing I’ve overlooked, but the first one might be something more complex…

You posted the same image twice.

It makes sense that the lines move like that. See my previous post.

Oops, the second image was meant to be this one, sorry:

It does make sense that the shadow moves down, as it would if the character was jumping, etc, but how to make it not behave like that…?

Quadmaps work better when the bottom of the drawing is flat where it meets the simulated ground. With those rounded edges you will not be able to simulate the way each one is at varying distances from the ground.

You will need to make the feet sit flat on the ground rather than at that perspective. The only other option I can think of would be to define the physics of your animated world and make all shadows stylized rather than truer to life. You could make them extra blurry oval clouds in every case. Just like The Simpsons have yellow skin and three fingers with a thumb, your world would cast cloudy blurry oval shadows.

Oh, yeah, don’t worry about the shadows’ look, I’ve made them blurry enough not to look bad.

The only thing I need is to make it move along with the character as he goes up and down… =\

To clarify this previous statement:

Quadmaps work better when the bottom of the object is STRAIGHT / LEVEL / FLAT where it meets the simulated ground. With those rounded edges on the bottom of the plant you will not be able to simulate the way each one is at varying distances from the ground.

You won’t have to worry about the feet and legs reflecting accurately if you blur the shadow edges enough.

But the staircase staggered walk cycle will never match in motion at that perspective angle.

I guess I’ll have to add a peg to the quadmap and make it follow the character up and down manually, if there’s no way to synchronize their Y position… Unless there’s something else I’m missing. =\

I am thinking it is more than that.

I think the objects have to have an edge with certain characteristics for a quadmap to look correct. The clue to whether a quadmap will work is that in order to match its source move for move the two must reflect one another and share the same point where they meet the ground plane.

Straight on perpendicular you can have an exact reflection. If you look down from above the reflection is not the exact opposite, it is from under the object reflecting areas that are blocked by the object. For instance, you might see the underside of a shoe reflected when looking at it from somewhere above.

Print out that green character and go in the restroom and hold it against a mirror so the feet re touching the mirror surface. You will see how the reflected version is not the same thing that you would get by taking that drawing then flipping and distorting it as a quadmap or in Photoshop. You have to be able to produce an object’s reflection by flipping and distorting it. If the pose requires something besides that it will not look right with a quadmap.