Attempt At Modeling A Real Person

[RS3D]

I have attempted to model a real person. You can see the head only on the first .MOV and then the same model with a torso, etc. in the second .MOV. I modeled the head, hair, and props, and borrowed liberally for the torso from a model called Marie Claire/Sister M that was donated by Klaus-Uwe Schomann. Thanks Klaus. I was not concerned with realism at this stage. I just want to see if the person is recognizable

before I go in and clean up the modeling and decals. Does anyone know who it is? I really compressed the size of the video, so I hope it is good enough.

F2_2.mov

F1_2.mov

[robcat2075]

The head-only looked vaguely familiar but I'm not a follower of that movie series so i couldn't place it.

 

I presume it's female after seeing the second movie but the face looks strangely male for some reason.

[markw]

I think the wand is a clue.

A younger Hermione Granger me thinks?

[RS3D]

It is supposed to be Emma Watson who played Hermione in the Harry Potter series. The problem in modeling someone like her is that pictures you can use for rotoscopes cover the time from about 9 yrs old to early teens. She obviously changed during that period so you have to just pick one. Attached are two of the pictures that I used.

[Rodney]

Nicely done. It looks like her to me.

[mborok]

I recognized her from just the head.

[ludo_si]

Wow, that's great, a beautiful model, I also recognized the character immediately.

[RS3D]

Thanks to robcat2075, markw, Rodney, mborok and ludo_si for your comments. I will keep working on the model to try to complete it.

I have worked on the eyes, cleaned up some of the splinage and located a really nice left side picture for rotoscope (see attached).

I have also made another .mov of just the head that I hope is better than the first one.

F2_3.mov

[robcat2075]

hmmm... comparing the model and the side reference... the point at the top of the nose that is farthest in appears higher on her than on your model and the chin on your version seems a bit flattish. Those are two things that stick out to me.

 

Be careful of how you copy that side pic since it's not exactly from the side.

[RS3D]

Thanks Robcat. This is the stage where I start to get nervous. Once you get a likeness, you run the risk of losing the intangible details that cause

people to recognize the character. You can wind up with a more accurate model in terms of geometry, but the model loses the intangibles that cause

recognition. I fixed up the two rotoscope images and placed them side by side with a profile of the model as it exists now. Then I superimposed

the Hash model profile on the rotoscope. It is not real obvious from these images what needs to be done, but it appears I need a better transition from

side of nose to middle of the cheek, move the bottom of the nose forward a little and experiment with the area between the eyebrows to move it

forward a little. That little protrusion between the bottom of the forehead and top of the nose is kind of unique to her and it appears in a lot of other

profile pictures of her.

 

Any and all suggestions are welcomed.

 

[robcat2075]

I've made several attempts at a realistic likeness of a known person and while the results looked human they didn't really capture the person enough to be obviously that person.

 

It is hard to do.

[RS3D]

I made a few more changes to the model. I also located a photograph that shows a 1/2-3/4 shot of the head. I tried to come as close as possible in

a choreography and came up with a comparison. I think I did pretty good on the mouth, forehead and eyebrow positions but it looks like I have to

carefully shorten the nose some more. The eyes are just blanked out in Photoshop because the model doesn't have eyelids yet.

[robcat2075]

Again I'd point out how the nose, from the tip to the eyebrow line, is not one curve but about three different ones along the way.

[RS3D]

Here is another update on the Emma Watson model. I put some specular on the eyes and shortened up the area from top of head to bottom of the nose.

I also changed the nose a little to get more of a button nose effect in order to get the multiple angles as suggested by RobCat.

Attached is a front face comparison between the original rotoscope (adjusted for hair style) and a frame from the choreography.

It looks like I am getting close, but I am still concerned that making it look like the rotoscope may not get someone to recognize the person.

I made another .MOV for review, so if anyone sees I am straying from "person recognition", please let me know.

F2_3a.mov

[robcat2075]

I still think the chin shape is departing from the reference.

[Rodney]

I like your latest improvements.

 

I'm new at color correction but I think that alone will help to sell a more realistic look for her.

Here's a quick take on adjusting the color... I also added a little film grain... not sure if that helped much.

I tried to keep your original on the right exactly the same but some changes might have crept into the final composite.

The color adjustment did blow out the hightlights a bit too much for 'realism' but I was trying to soften the skin a little as well.

Perhaps that has moved my adjustment a little further away from the realism you are after?

 

I"d say the biggest difference remaining between your two images might be her upper eyelashes... or lack of them in your latest improvement.

 

 

F2_3a (color correcting comparison) sm.mov

[Rodney]

I will add that if you are using your rotoscope only from the bottom (i.e. splining over the top of it) then you might want to place it on top with the transparency turned up.

Then as you adjust your CPs and splines you can better see where they are off from the original reference.

While all features are important that triangle between eyes and nose is the area to focus on.

Within that area I'd say the eyelashes and tip of nose (from the front view) are in need of attention.

The eyelashes appear to be largely missing and the nose should be a little more bulbous.

 

Hope all of that makes sense.

 

 

Edit; Some of the images you've posted suggest you are using the method I describe above.

[robcat2075]

I'll note another challenge to using most photos as rotoscopes is that they have a perspective viewpoint that is not impossible but difficult to recreate in our modeling environment.

[Rodney]

I'll note another challenge to using most photos as rotoscopes is that they have a perspective viewpoint that is not impossible but difficult to recreate in our modeling environment.

 

 

Robert makes an excellent point and I was tempted to mention a good workaround to the problem of rotoscopes problem in my previous post.

I tend not to use the default rotoscope... or at least avoid it wherever possible.

I prefer to use a single patch with an image (or image sequence) applied to it.

In this way I can manipulate the 'roto' in just about anyway I need to (rotation, scale, skewing in any dimension...)

To go one step further one could apply a 'roto' image (or sequence) to a grid as a decal and then distort that grid as necessary.

This method of distortion is even more useful if caricaturing a subject. You can get the basic features right but exaggerate via the grids distortion.

 

Other software often relies on setting up two to four rotoscopes... one for each of the cardinal directions but it's even better to be able to capture those 3/4 views that a single image rotated in space can convey for you.

 

In A:M, yet another alternative to the classic rotoscope would be to use a Layer, which can be rotated in any direction, but if I'm going to go that far I usually find that a patch image will work even better.

Layers have the ability to adjust some surface properties that 'rotoscopes' do not... such as cranking up the ambiance setting so that the finer details (the ones that are less important) are removed.

Patch images have a whole lot of other options and can be manipulated via poses, etc.

 

Specifically to Robert's cautionary note is the fact that one can recreate the general perspective more easily by allowing for the view of the camera.

If a sequence of images roughly the same size can be strung together and then the view adjusted while creating the model.

 

But this doesn't address the issue of the 'error' the camera perspective produced in the first place when it captured the image so...

This leads me to the ultimate suggestion which would be to try to use a sequence of images for reference.

I'm not suggesting anyone slavishly use this reference footage to trace from but instead use it to capture the feel of the character in movement.

 

And.. all this advice is from someone who prefers cartooning because 'realism' is too much work!

[robcat2075]

If one knew the exact vantage point from which a photo was taken, the distance, the angle, etc., one could recreate that with a camera in the chor, put the rotoscope on that camera and model from that view. you could do that with any number of oddly angled reference photos if... if you knew their vantage points.

 

Trying to deduce that from just the photo itself is nearly impossible, I regret..

[RS3D]

Thanks to robcat and Rodney for your suggestions and observations. I really appreciate it. I will try to respond:

 

Rodney says

[i"d say the biggest difference remaining between your two images might be her upper eyelashes... or lack of them in your latest improvement.

While all features are important that triangle between eyes and nose is the area to focus on.

Within that area I'd say the eyelashes and tip of nose (from the front view) are in need of attention.

The eyelashes appear to be largely missing and the nose should be a little more bulbous.]

--------------------------------------

All valid points. At this time the model does not have eyelids and putting them in is going to be tricky. I am saving that for the last stages where I can test eye blinks in 3d. The nose issue is also tricky. As I make it wider in the front view, it tends to lose some of the curvature in the side and 3/4 view, so I am going to have to do changes in little bite size increments.

====================================

Rodney says

[i will add that if you are using your rotoscope only from the bottom (i.e. splining over the top of it) then you might want to place it on top with the transparency turned up.

Then as you adjust your CPs and splines you can better see where they are off from the original reference.

--------------------------------------

I put the rotoscope down first and then model over it. I use the rotoscope image also as the front face decal and I can align it perfectly with the underlying rotoscope decal. This produces some distortions because I am not using a flattening procedure before applying the decal. I then go in and adjust the splinage to minimize undesirable effects]

=================================

Rodney / Robcat says

[in A:M, yet another alternative to the classic rotoscope would be to use a Layer, which can be rotated in any direction, but if I'm going to go that far I usually find that a patch image will work even better.

Layers have the ability to adjust some surface properties that 'rotoscopes' do not... such as cranking up the ambiance setting so that the finer details (the ones that are less important) are removed.

Patch images have a whole lot of other options and can be manipulated via poses, etc.

Specifically to Robert's cautionary note is the fact that one can recreate the general perspective more easily by allowing for the view of the camera.

If a sequence of images roughly the same size can be strung together and then the view adjusted while creating the model.

But this doesn't address the issue of the 'error' the camera perspective produced in the first place when it captured the image so...

This leads me to the ultimate suggestion which would be to try to use a sequence of images for reference.

I'm not suggesting anyone slavishly use this reference footage to trace from but instead use it to capture the feel of the character in movement.]

------------------------------------

I am basically trying to do the things you describe in slightly different ways. Mainly by manipulating various settings in ambient lighting, image based lighting, etc. in choreography. I agree that the best way to do this project is by trying to keep several images from the movies in your head rather than trying to duplicate one image. That is why I started with a 3d animation rather than a specific image in the first post to the forum.

====================================

 

robcat says

[i still think the chin shape is departing from the reference.]

----------------------

Which angle is giving you the problem? is the the side image or the front image?

[robcat2075]

In defense of slavishly tracing... that's pretty much what they do when they 3D scan a person's face or plaster cast of them. That's like the ultimate of slavishnessy behavior.

 

If one could recreate the original vantage point of a reference photo then you'd be neutralizing the perspective distortion of the photo.

 

 

On the issue of the chin... i notice that the reference photo appears to be from eye-level while your render appears to be looking from fore-head or crown-level. That complicates the comparison.

[RS3D]

Thanks robcat. I will watch for chin issues in the animation playbacks.

Attached is an "eye candy" animation to see if the decals will hold up under different lighting conditions.

 

F2_4.mov

[RS3D]

Here is another update to the model. It now has jaw movement, teeth, a frown pose and hair mobility. The animation shows typical turnaround, mouthing the word "hi" followed by hair swishing. I hope the changes I made do not reduce the recognition factor. So if someone sees a problem please let me know.

F3_1212.mov

[RS3D]

I am trying to put together an animation for the Hermione character. Hopefully it will be a spectacular feat of magic. But before I get too far along on it, I would like people to tell me if they have any problem with the layout of the scene-such as perspective, lighting, prop positioning, etc. Attached is a teaser that shows the environment for the action.

candleOnly.mov

[robcat2075]

I am trying to put together an animation for the Hermione character. Hopefully it will be a spectacular feat of magic. But before I get too far along on it, I would like people to tell me if they have any problem with the layout of the scene-such as perspective, lighting, prop positioning, etc. Attached is a teaser that shows the environment for the action.

Let's see... just brainstorming here... Does the table need to be that big? Can it be smaller so you can move in closer to her and the props on it?

[RS3D]

Thanks Robcat for your input. I was mainly concerned about the relative size of the girl model to the table. She is supposed to be between 10-12 years old, so the table size is an attempt to help scale the model in the scene. I stood next to a table today and tried to figure where the table would be with the girl in this scene. I added reflection to the table and that seemed to help integrate the model into the scene.

 

I actually decided the girl model is too big, so I scaled the girl model and the candle down and re-positioned the camera (see attached). The width of the table will be necessary for the eventual action that will take place, so either I start closer and then move back of just leave the camera alone. I appreciate your help.

[RS3D]

I spent time to clean up the model to make animation easier. I am going to try for about 15 seconds worth of animation. Four second sequence of book summoning -see attached (VERY rough) .MOV. Then

3 seconds of book page reading with different camera angle, 4 seconds of candle action and then 4-5 second payoff. If someone could look at the opening animation and give me you impressions related to timing and pose selection, I would appreciate it.

BookOnlyMove.mov

[RS3D]

I added a few things to the opening 6 seconds and tried a larger, more complete animation.

BookOnlyMove2.mov

[robcat2075]

is that the way they hold the wand in the movies?

 

i think of the way Tinkerbell would hold a wand as if she was going to tap things with it, that just says "magic wand" more.

[robcat2075]

The most significant "animator" problem I see is that you have lots of "isolated motion"

 

Her arms are moving independently while her torso remains locked in space. It's just about impossible to move one's arms without some effect on the torso and typically one would turn the torso into the gesture to put some shoulder into the motion toward the book.

[RS3D]

I made another update to the book summoning section. Following Robcat's advice, I put more secondary and directed movement in the character. I checked a Harry Potter movie and they use the wand as sort of a weapon or in a kind of violent way. They usually have a pretty good grip on it. The hand model kind of bothers me because it looks more like a blacksmith's hand than a young girl. I will have to redo that part. This update includes the second camera action: viewing the book from behind the character. I allocated 80 frames to that part so hopefully it is enough time for the viewer to read the designated page in the book.

BookOnlyMove3.mov

[RS3D]

Here is another update on the animation project. I worked mostly on getting the various props, effects and camera locked down to fit the time budget for each segment. I still need to replace the right hand/wand combination and do a lot more work on secondary motion, anticipation, etc. You really have to concentrate on key framing everything in a project like this. So far, I have done it all in one choreography.

BookOnlyMove4.mov

[robcat2075]

Have you tried acting this out to a camera and watching that for reference? That would probably tell you a lot about the timing and size of motions.

 

The initial gesture with the wand toward the book i find confusing because It's not clear which of the strokes is the magic one.

 

 

There is a great Keith Lango tut on "snap"... gesturing with an arm... that would serve you well on her left-hand gesture with the open palm. Right now you have just the hand rotating on the end of the arm but it can be made much more effective if the whole arm contributed to the motion.

[RS3D]

I completed blocking in the different scenes for the total animation and now I am going back and tweaking the individual actions. Attached is the latest book summoning scene where I tried to implement some of RobCats suggestions along with stronger poses. Any feedback will be appreciated.

BookOnlyMove6.mov

[RS3D]

Here is another update to my animation. This one has the entire animation (27 seconds) with a sound track added. I took out the reflections, shadows, & background and carpet images to get a better focus on just the animation in the various scenes. I wish I knew more about using the camera for highlighting various aspects of this animation. So much to learn, so little time.

HermioneLesson.mov

[RS3D]

Here is a final-full rendering version of the animation. It is very compressed (77meg down to 11meg), and the quality suffered a little, but it seems to be viewable. Thanks again to everyone who responded with suggestions, feedback, etc.

Book0_830a.mov

[robcat2075]

I think that's a fine first expedition into animation you got put together there. There's a lot of complicated manuevers in the scene.

 

If the text in the book was importabnt for us to read that probably could be substantially bigger on the page, but I think I get it even without that.

 

Idealy, there would be something to make that first candle fall, like maybe she accidentally nicked it with her wand as she's looking at something else.

[RS3D]

Thanks for the input Robcat. The collapse of the book and candles together is supposed to show that her whole lesson is not working because of her being a novice who is just

learning the necessary skills. As for the reading times, I agree, but I think the whole animation feels a little rushed. I could go on making improvements in just about every area including

modeling, lighting, animation. etc.

 

It is definitely not a monument to good animation, but I accomplished what I set out to do and I learned a lot about putting together a whole mini-story instead of just a scene.

As Rodney pointed out, working with real characters is a lot of work, but I also think it is a lot of fun. I am going to work on something a little easier next like doing a semi-truck model.

[robcat2075]

Onward and upward! Or sideways!

 

Whatever the next interesting project may be!