"Look At" - eyes offset [Solved, I think]

rames44

Has anybody else run into this problem?  I've seen it with both G3M and G3F.  If you set a character's eyes to "Look At" an object, the don't really appear to be pointed there.  The left eye seems pretty close, but the right eye is offset.

I've attached a screen shot to indicate what I mean.  As you can (hopefully) see, I've created a Null in this scene, and I've positioned it exactly at the same spot that the main camera is located.  (Yes, I know I could have used the camera, but I was experimenting with moving the "eye target" around.)  Both the left eye and the right eye have been set to "Look At" the null.  But, as you can see from the render, the right eye isn't looking there - it's rotated too far to the figure's right.  The left eye seems pretty good.  (Try holding your hand over one half of the face and then the other and decide where the eye you can see is pointing.)

As I've said, I've seen this with G3F as well.  It's almost like the pupil and iris on the right eye isn't really centered on wherever the (0,0,0) point of the eyeball is.

Either that, or "Look At" isn't calculating the rotations correctly.

mindsong

I generally agree with your assessment, as it often feels like characters set to look at the camera are looking through you, rather than at you.

Without testing G2/G3/G8 (I use G1 but see same effect, as well as with the Mil4 series).

I wonder if you move the target extreme-ly, if both eyes follow reasonably well, or if you get a lazy eye... If one eye is lazy, perhaps both eyes need to be set to track the item. (probably obvious and you're already doing that, but thinking out loud here...)

Also, as you mention... I wonder if the eyeballs are considered centered on their own core, or from another pivot.

Something ain't right, to my eye.

I'm also going to be dealing with this in the near future (testing and rough-drafting right now), and was thinking I might point at a null and try moving it back beyond what would seem like the intended target point, or perhaps closer to the figure than the target point, and see if that would make a difference.

Dreamlight (PA at DAZ) has a product/tutorial on eye-contact in the DAZ store, but the price was too high for my interest, so I've deferred buying it until all of my theories are proven wrong...

let us know what you learn!

cheers,

--ms

Llynara

I set the eyes to point at the camera all the time, and it used to work fine, though sometimes you had to move the camera slightly to get the eyes to follow. Lately I'm having more trouble with it. I definitely notice this issue on older figures (my render last week with V3 had this problem.) But I'm starting to notice it with G3F as well. I recently did a book cover and one of the eyes was really skewed. I ended up having to adjust everything manually and it took a long time to get it right. Not sure what has changed, but I don't remember having this problem before the updates came out around G8F's release.

Fletcher

I've been seeing this problem latwly and i'm not using cameras. It's must be some rogue morph. For the original poster, I see your camera is very close to the subject. You can try moving the null very far away beyond the camera and have the eyes point to that. This may solve a bit of the crossed eye.

rames44

@Llynara I first ran into this problem before the G8 updates, but don't recall it when I first started using Daz. If it's a rogue morph, I wonder how one would identify it?

@Fletcher Yes, moving the null farther away does lessen the effect in the scene, of course. I set up this scene to illustrate the issue. 

I also notice, oddly, that the eyes pick up a bit of a twist as part of the "look at," not that that really affects the render. 

Fletcher

Rogue morphs are a pain. You have to keep removing the latest morphs until the problem disappears. In your case, it's a morph that affects both G3F and G3M so that narrows it down. But to verify it is a rogue morph, you have to use the same scene setup on a clean figure with just the basic DAZ morphs and see if you have the same problem, if you do, then it's not a rogue morph.

rames44

Fletcher said:

Rogue morphs are a pain. You have to keep removing the latest morphs until the problem disappears. In your case, it's a morph that affects both G3F and G3M so that narrows it down. But to verify it is a rogue morph, you have to use the same scene setup on a clean figure with just the basic DAZ morphs and see if you have the same problem, if you do, then it's not a rogue morph.

I'm actually in the process of downloading a virgin Daz onto a different computer, with just the starter stuff to see how that behaves.

If it is a rogue morph, would it be referenced inside the DUF file?  Or is this (possibly) something that's essentially been "installed" into the innards of Daz that gets invoked without leaving any traces in my project?

Fletcher

It's not as complicated as that. It's just that is so tedious to find it. What you would do is remove your the morphs from your most recent purchased item, it's in the morph directory of your data directory, i don't know the exact path at the moment, refresh your library, test out your scene. If the problem goes away, you've found the culprit. If not, you repeat with the next purchased item. It's best to test it out with your clean build before going through all this.

rames44

After a little experimentation, I think I've come to the conclusion that this effect is actually due to the way that the model (Ivan) is designed, creating a bit of an optical illusion.

In order to eliminate any possibility of rogue morphs:

1. Install a virgin copy of Daz on another computer I have.
2. Download only the stuff that comes with DAZ out of the box - starter essentials, etc.  No purchased "stuff."
3. Drop the raw G3M into a scene along with a camera, and set both eyes to "Look At" the camera.
4. Arrange the camera to get a closeup of the face by positioning it and then fiddling with the focal length.  Image attached.
5. Move the camera around and see what it looked like.

When I did this, the on-screen image looked pretty good.  As I rotated the camera around (within reason), it appeared that the eyes were tracking me pretty well.  So this seems to indicate that G3M is behaving as expected.  (I assumed as much, however, in those famous words of software testers: "trust, but verify.")

Step 2, try to recreate the problem with the minimum of add-ons:

1. Add just Ivan (and Ivan HD, since that was used in my original) into DAZ.
2. Open the scene without making any changes.  Fiddle with things.  Everything seems unchanged.  So adding Ivan into DAZ didn't do anything gross.
3. Now, without altering anything else in the scene, dial up the Ivan head.  (Image attached).
4. Then dial up the "smile" posing control.  (Image attached.)

Because of the way I did it, the eyes and head didn't change position as part of this operation, yet after Step 3, Ivan's right eye appears (to me) to be slightly off in the second shot.  Following Step 4, it appears to be significantly off.  Yet the actual position of the eyes did not move at all.  What moved was the facial features around the eyes.

I have a feeling that this is an illusion caused by the fact that Ivan's eyes (or, more correctly, the opening created between his eyelids) are narrower and "squiny-er" than the raw G3M's, and this, in turn, creates the illusion that his right eye is pointing farther to the right than it otherwise really is.  Then when you dial up the "smile" expression, the eyelids are again shifted a bit, exaggerating, I think, this effect.

In real humans, as the eyes move back and forth, the eyelids and the opening they create also change shape somewhat because of the corneal bulge and because of the way our brain controls things.  For example, as we look up and down, our eyelids automatically adjust up and down somewhat.  This doesn't happen with these figures - they're not really that smart.  (Try doing what I did with G3M and then rotating the camera down so that it's at his waist level or well above his head - looks very creepy because the eyelids don't move.)

So I think I've concluded that, in my original shot, the eyes may indeed be pointing in exactly the correct direction, but they just don't LOOK like they are because of, well, "reasons." LOL  I can obviously correct this by parenting the eyes to two different nulls and adjusting them separately until things look right to me, but it's an interesting exercise in how subtle our brains are in interpreting things from what we see.  Our visual systems clearly use the position of someone's eyes with respect to their eyelids in very subtle ways.  (And of course, with a real person in front of us, noting is ever completely static they way it is with CGI)

Also, as a parenthetical - when I look at the thumbnails of the images I'm uploading as I post this, the eyes don't appear to be pointing in exactly the same position they appear when the images are full-sized.  When it comes to the eyes, what a difference a tiny fraction of a pixel makes... :)

mindsong

You could probably get into the 'dominant eye' effect, where the head and dominant eye will always determine a straight-on look, with that eye centered, and the non-dominant eye will be slightly off-center as a natural result. I wonder if' we're that tuned in to 'norms' that when its not quite right we end up amplifying the uncanny valley effect.

you're right about the thumbnails vs full-sized too. interesting.

--ms

whitefire

Hi

I believe this is not a problem with DAZ Studio, really. It has to do with what it actually means to 'look at someone'. You see, When you have another person looking at YOU, they are NOT really looking at the surface of your single eye (left or right). Test it out, let someone look at your eye, not at you. Feels weird, huh? When you really look AT someone, you are actually looking into some infinity behind them. And no, I am not preaching any religion here :) Just physics. So, in order to deal with it in DAZ I just set up two cameras or objects separately for each eye to look at and set them up separately. Weird solution, lot of work, but you can absolutely deal with the problem this way.

leka7575

Hi

whitefire you're right! Everything works that way. To save a character from strabismus, I create a primitive (small sphere) behind the camera at a distance of about 400 axes, then selecting both eyes tie them to the sphere ("Persistently Point...") I attach the sphere to the camera (simply by dragging the camera onto the scene menu).