Question Look-a-likes with custom expression morphs?

[btacoses]

I am or have been a patron for a ton of creators on the hub and have a gigantic collection of looks. I've tended to have a preference to celeb look-a-likes, even for celebs I don't know. They tend to have a lot more realism in the face even if the likeness isn't perfect. That facial realism makes a huge difference for me. But here's what I don't think any of these looks have, custom expression morphs to animate the face of a look-a-like to go from expressionless to a common expression of that specific person. I've seen a couple looks posed with expressions that are close. In @vecterror's 3 Angels for VAM, two of the expressions saved in the look are very accurate. I tend to use @ClockwiseSilver's random expression plug in, and it would be great to chuck one or two custom expressions in with some really subtle facial expressions like brow raises and smirks. The fact that this isn't being done frequently makes me think this would be too difficult, but I don't know why it would be. So I guess that's my question. How hard would this be? And also I guess another question is more speculative, why isn't this being done more frequently?

 

[TToby]

I guess it is the way you would define that "typical expression".
Maybe some creators don't realize it to be typical in the same way like you.
Some creators and users may even prefer a neutral expression, for using expression-plugins for instance, or for using their own expressions.
Think of a celeb with a baked-in fat smile: If you will use a plugin, or some poses with expressions, they sometimes will look awful and horrible.

The IMHO best solution for you could be, to learn how to create your own facial expressions.
If you don't allready have, I would strongly suggest to download some of those expression-packs, available here at the hub.
Some of them are modular, so you can almost create every expression you could think of.
Once you are getting used to it, it is very simple and a matter of only some minutes. Maybe this is why some creators do not care that much about expressions.
VaM is a sandbox tool. It is a dream for tinkerers and creators, but sometimes can be a nightmare for those who are not.

 

[btacoses]

I guess it is the way you would define that "typical expression".
Maybe some creators don't realize it to be typical in the same way like you.
Some creators and users may even prefer a neutral expression, for using expression-plugins for instance, or for using their own expressions.
Think of a celeb with a baked-in fat smile: If you will use a plugin, or some poses with expressions, they sometimes will look awful and horrible.

The IMHO best solution for you could be, to learn how to create your own facial expressions.
If you don't allready have, I would strongly suggest to download some of those expression-packs, available here at the hub.
Some of them are modular, so you can almost create every expression you could think of.
Once you are getting used to it, it is very simple and a matter of only some minutes. Maybe this is why some creators do not care that much about expressions.
VaM is a sandbox tool. It is a dream for tinkerers and creators, but sometimes can be a nightmare for those who are not.

I am not suggesting baked in expressions. Those are the worst. What I'm suggesting is packaging the expression morph in with the VAR instead of merging all the morphs into one. Merging facial expressions into their own morph that is separate from the morphs necessary for the look.

I have created a preset for the expression randomizer that uses some of the very specific and subtle face morphs that works really well on a couple look-a-likes, so I know what mean about creating them. But when a designer is working with reference images it would be so much simpler to do it at that stage. I've tried moving reference images into VAM for this purpose, but it takes so much time for me to set them up to work right. A designer has likely already done that work to create the look.

 

[TToby]

Hmm... OK, I think I understand you now.
Merging only the expressions would be a possible way, instead of linking to the different expression packages as dependencies. Many users hate unnecessary dependencies, but having 10 versions of the same morphs in different Var would be worse. But I think it may be too much hassle for many creators. Many of them doesn't even optimize the Var linkage or the texture files size.

For setting up referenz images:
I would suggest to create your own starting scene by saving one with the name "default".
Or at least save an empty scene with a image panel in it.
Because I create looks most of the time, I have a starting scene with an empty image panel in the right orientation behind the figure. If I need a reference picture, I simply click on that panel and load one from a certain "reference" folder. If I don't need one, I simply ignore or delete the panel.

 

[DJ]

You may run into the same problem i had when i tried to make a few nice smile morphs: What looks great on one face does not look good on another. Good smile morphs seem to be look-specific.

It takes some time to really nail down a nice expression. That's a lot of work for very little return if the morph only end up being useful on the one look.

 

[TToby]

What many of those celebritie-looks make look so real, are IMHO most of the time face textures generated with FaceGen from photos. I personally like them very much, too. To create one of this, you need a very good photo. It is very difficult to find high resolution pictures, that are usable. But it is even more difficult to find some with only a neutral expression.
All those generated face textures looks so life-like because of some skin-details, but also because of shadows, wrinkles, make-up, facial hair, highlighted parts, aso, baked into the texture.
So, none of those generated faces are really "empty" like those DAZ3d standard textures coming with VaM.
The human face has approximately 50 muscles and you even need 17 for a smile. Not to speak from all those different wrinkles.

I think this is the reason for what @DJ said:
If you create a facial expression on one of those faces, that looks good and "right", it may look completely off by only changing the face texture! On the other hand, if you have good expressions for a standard "empty" face, there is a high chance that it doesn't work on a generated face.
I would even go so far to say: If you have a generated face texture by FaceGen with a baked-in expression, only those expression plus some similar ones will really work for that texture. With some exceptions, like always.

Putting it to the extreme: a good look-alike creator will most likely add a face expression that fits best to the face texture he has generated, because otherwise it wouldn't look "right" and therefore "recognizable".
In this case, there allready IS a "typical" facial expression, like requested by @btacoses .

If the creator doesn't, it would be one of those cases with a great product-screenshot, but very dissapointing when you load that look... (with one exeption: some of those face textures works even without any expression-morphs. I call them "textures with baked-in expressions").

BUT, and that may be what the OP said, those included "typical" expression may only be visible in the included scene. If you extract the look you have bought and put it in your own scenes, the expression may be resettet (if they are correctly marked as "pose") and the face doesn't look that much "real", like before.

To solve this issue, there are IMHO three ways:
-To bake-in a specific expression into the full-body morph, which can cause issues with other expressions e.g. from plugins.
-To create a face texture that works with each and every expression you will throw at it... very difficult, see above.
-Or to merge the expression-morphs used in the preview-scene to a separate and very specific morph, that maybe only looks good on that single look.... I think I never have ever seen a look which has included one of these?

You may see, this isn't as simple as it seems on the first sight.

 

[btacoses]

-Or to merge the expression-morphs used in the preview-scene to a separate and very specific morph, that maybe only looks good on that single look.... I think I never have ever seen a look which has included one of these?

This is what I'm trying to explain right here. For most look-a-likes, there is a single merged morph for that look. What I'm suggesting would be like merging all the morphs with the face as neutral as possible, and then merging only the morphs used to turn the face from neutral to a slight smile or smirk or whatever we often see the person make in photos, etc. There's a Sara Silberman look-a-like who I have managed to almost perfectly nail her usual slightly-smirky smile. I have merged all those expressions into one morph that I animate with timeline. Then I use expression randomizer with the same 20 or so morphs I used to make the one merged expression, set with a fairly narrow and low min/max settings. With both of these plug ins running, I have managed to not only animate the face but to make the look appear much more like the person. Maybe I should credit my skills more and realize that it wouldn't go much faster for experienced creators than it did for me, in which case, you're right, it would add more time than I originally assumed for the creators to do this.