RS3D
-
Posts
142 -
Joined
-
Last visited
-
Days Won
2
Content Type
Profiles
Forums
Events
Posts posted by RS3D
-
-
That was fun to watch. I admire anybody who works through a whole project like this. The last few seconds when the character is winding down the song were really great.
-
Excellent video. It should make setting up the key frames easier. Any front views available?
-
robcat2075 Posted Sep 14 2012, 07:45 PM
hmmm... the girl and thom don't seem to be casting any shadows on themselves even though the light appears to be from nearly overhead. That's what i notice
----------------------------------------------------------------
I tried to create more contrast by changing the top down lighting to be stronger and decreased the bounce lights that I used to light the front of the models. This results
in loss of some detail in the image characters but seems to make the scene more realistic.
Rodney Posted Sep 14 2012, 07:54 PM
The voice in my head tells me to say...
It would be nice to gain a little more feeling of depth than what we are seeing here.
Compositionally this would have the background fading to a lighter hue while the foreground would be darker.
--------------------------------------------------------------------------------
I de-saturated the top half of the background image and blurred it a little. It did seem to add depth to the scene.
Is this one frame of a sequence?
---------------------------------------------------------
It is one frame from an animation. But before I re-render it, I wanted to get more realism in the choreography.
Added: A trick I have used to get different elements not from the same source to match better is to add noise over the top of everything.
------------------------------------------------------
I tried different hazy materials on a front plane, but the changes in lighting seemed to be enough to get the characters to appear to merge with the background.
I will look up the source you cited and try some other changes when I get some time.
NancyGormezano Posted Sep 14 2012, 08:26 PM
For me, the self shadowing on Thom (and other character) is too soft, compared to the dark, sharp shadows on the plants. And it does not appear that Thom is casting any shadow on the ground. The self shadowing also seems to be "wrong color".
For my taste, I would also say that the color of Thom is not as saturated as the colors in the background image, so it looks odd, unbalanced. Something to try to make the colors look more consistent: use Image based lighting, and use the background image as your source. The colors on the models would have a better chance of blending in (would have to try to see if so).
EDIT: This may not suit your taste, but I made the global ambiance color more orangey, to make Thom more match the bright sunflowers in the bright sunlight, and the shadow color on the flowers is more consistent with Thoms self shadowing. We now have Thom with dark black shadows on the ground.
The ground is flat shaded, front projected & there is only 1 white klieg overhead casting shadows.
It could use some tweaking. I tried using your image for the Global IBL, but it didn't do what I was thinking it would do. I thought the orange global color worked better (for my taste).
-------------------------------------------------------------
I tried using the image based lighting, but I kind of gave up trying to get a better effect. I'm sure it is a good way to go, but I just do not have the knowledge to make it work.
I did use the global IBL with an orange color. This, combined with the lighting changes kind of made everything blend better. I still would like to get the effects you suggested
with the shadows on the Thom character, I tried different processes, but again I do not have the knowledge to pull if off effectively.
==================
I think this new image is better (more realistic), but I seemed to have lost too much detail in the characters because of the shadows. I think I will have to work on some of
Nancy's suggestions for changing the lighting characteristics in the shadows to restore some of the detail.
-
That was really good. I especially liked the little subtle face expressions by the band members. Things like that set it apart. It must have been a lot of work. Animations that take
30 seconds seem like a lot and yours is over 2 1/2 minutes.
-
Thanks for the excellent suggestions. I will work on them this coming week and post the results.
-
I have been working on getting the lighting correct for compositing Hash characters into photo backgrounds. This can be tricky sometimes because you have to take the entire scene into account in selecting the lighting. Please look at the attached image and see if the characters appear to be be part of the background image.
-
oops. I forgot to say thanks for posting that site. There is goldmine of lighting ideas in those images.
-
-
Awesome! Very clear and easy to follow. Thanks a lot.
-
Thanks for posting those. I will play with them when I get some time.
-
When people move their head to music it seems as though they move it more forward and back (head bobbing) to the music. Yours is more side to side.
Maybe you could try to get it going forward and back and see how that looks.
-
Looks a lot better. Are you going to put in some face movements? With that music intro, I would like to see that he is serious about getting ready to rock and roll.
-
Something that might help is to exaggerate his movements more, especially his head. Since he is a cartoon character, you can really go big and it will still work.
-
-
One big thing in dealing with polygon based models is to look at the size of the model ( either 3DS or OBJ). If it is more than about 1.5 meg for OBJ or more than about
300-400 k for 3DS it is probably not worth the effort. I have converted a few of the OBJ models at around 1.4 meg in around 5-10 minutes. After you do that, you can look
for the parts that are the big space hogs like spheres and other curved surface items and copy them to another model, where you can work on them by using them as
as a rotoscope to help in replacing with the Hash primitives that will look the same and require about 1/100 the amount of space. By replacing (copy/paste) the original parts
with the new Hashified parts you can start to get the model down to a size where you can work on eliminating most of the triangles.
As far as textures go, sometimes the textures work with the converted model and sometimes they don't. It seems like around 50-50 chance they will work
automatically as a result of the conversion.
However, if you look at the groups section of a model, you will notice that the groups tend to be parts of the model that correspond to the textures that come with the model.
Then it is just a matter of importing the texture into the Hash project and reapplying the texture to the proper group just as you would for a standard Hash model.
I hope that is helpful.
-
I have been experimenting with some of the game models available at http://thefree3dmodels.com. There are over 1,000 models there, most are in OBJ and 3DS format.
The site is free and the models download pretty fast. Because many of these models were used in the game industry, they are set up to be efficient in terms of detail.
They are much smaller than the usual file sizes for these formats. It is interesting to see how they were put together to maximize the ease of use and texturing processes.
The included textures alone are worth examining the models.
I got interested in one of the face models because of its simplicity and the way it takes decals without too much distortion when used from various viewing angles.
I have attached some pictures showing some of my experiments. The first one shows 4 different faces, all made with the same face mesh. The 2nd, 3rd and 4th picture
show how one of the faces looks from various viewing angles. This model is just the face mesh (+ my own hair model), with a decal slapped on it. I also attached a MOV of the helicopter pilot showing how the model holds up in an animation situation.
I did spend a little time on the helicopter pilot model. It has a bump map and specular map in addition to the color map. I think I can get it even better with a little more
work on the decals.
-
I use your stuff all the time. Thank you and have a happy birthday.
-
RS3D
I am sure that you will understand the new things that you have to change in order to get better results, but as one of our colleges said it has no need to be perfect! (but I would tried it to be perfect if it was mine work
So I have attached two images. First, is the image showed on the first frames of the animation walk. The image is the one with the right foot at your right side of the screen (I called it "right foot in front"). You will notice with the help of the ruler that that position does not exist at all in a human walk-cycle (only in Animation Master
): the right foot is supposed to be the one firmly touching the floor but it is flouting on the air. And i can notice it when she is walking, like a subtle jump the second image is showing us the second inaccuracy of the walk cycle. Here the left foot is in front (on your right side of the screen). BTW, it does not matter which foot is in front. You have to fix both sides.
Continuing, if you watch your animation in slow motion and reach this part of the animation you will notice that the rear and the front foot are moving at the same time (one down, other up). To let it with a better look just, I assume, you have to delay a little bit the moving of the rear foot (right foot). Allow the front foot (in this case the left foot of the character) to touch the floor firmly before deciding to move up the other foot from the floor.
Only with trial and error you will find the exact delay time, unless experience user of A:M have a tip for accomplish that more easily. You may ask them.
And this is the best i can explain what I have been seeing.
right foot in front
left foot in front
These are excellent observations. You are correct about the problems in the walk cycle so far. However, the real problem is that the
legs never reach a fully extended position. If you look at almost all walk cycles, the front leg extends straight (or almost straight)
in the first frame and then again at the point where the hips are at their highest point (usually about half way through the cycle).
As Robcat's tutorial points out, the leg actually is straight a little past the half way point in the cycle. The problem I have with the
current model/bone rigging set up is that, when I keyframe the straight legs, the animation shows rather noticeable knee popping.
You can look at previous posts from Robcat explaining the need for either changing the rig to a more sophisticated one or
trying some other tricks to get to where I can use the straight leg position. Rather than changing the whole rig and model set up,
I have tried to come as close as possible to the straight leg position without causing the undesirable popping action. This means
you either raise the hips or raise the foot to try to compensate. Raising the hips looks worse than raising the foot.
What you see in the first frame is the cheating the problem by raising the foot off the ground by about a keyboard click. This
allows the hips to continue down for a couple of frames after the first frame. This helps give the feeling of weight to the
walk cycle.
The straight leg problem also comes up again in your second point, that the legs appear to be moving at the same time. Because
I have to use a bent leg position, I have to compromise between bent leg/flat foot and bent leg/bent foot positions to try to overcome it.
(I hope that is clear).
I am afraid at this point, I am looking for a believable walk (per Mouseman's post) rather than a perfect walk.
-
Hi
This last sample side of your walk cycle is much easier to follow cause the dimention.
First before I respond to your questions, check the dimentions of the legs. See it in your last side walk-cycle.
It may sound strange but if you draw an horizontal line that touchs the very end of the right foot when it slightly touchs the ground and then you draw it back to the left foot, you will notice that the line will touch inside the foot, as it has entered on the ground.
Please check it yourself in order to see if it is my impression or not.
I am not sure if this may help to solve it.
If I understand you correctly, are you saying that the rear foot in the first frame of the animation should rest on the toe rather than
more toward the middle of the rear foot?
-
RS3D,
I have being following your walk cycle and I have seeing something for you to think.
I believe that everything is more or less on their way, and that there is still something concerning the shoes (and its articulation, I do not know its name). They look too straight, too horizontal as they move, and there should be, I think this is the point, a slight movement with some inclination to the back at the time the shoe is leaving the ground.
Do you mean that the foot should bend more as the heel is lifted off the ground? I have attached a side view of the walk cycle. Please indicate
where in the video you think the problem occurs.
-
Here is another update to the walk cycle. This one uses more up and down movement of the shoulders that are the opposite of the
orientation of the hips. That is, as the right hip rises, the right shoulder lowers and I used the opposite sequence for the other side. It softens the
walk at bit and seems to make the model more feminine. I also added stride length control.
I used kinematic constraint to a null to control the head and it seemed to do the trick for this application, although I plan to use a suggestion from Robcat to see how it might work.
I also attached another small sample of a front view of the walk cycle showing the shoulder movements.
-
I forgot to ask if anyone knows a way to lock the head bone position while moving the upper back bones.
You really mean the orientation, right? The head bone is attached to the neck and spine and has to move with them.
you can isolate the rotation of the head (or probably better, the neck) by Orient Like constraining it to something higher in the hierarchy than the spine hones. The first spine bone is a likely target.
I tried adding a null to the bone list and constraining the head to it as a kinematic type of constraint. This seemed to do the trick for what I wanted
to accomplish with the shoulder movements conflicting with a relatively steady head position. I will try your suggestion also.
-
I forgot to ask if anyone knows a way to lock the head bone position while moving the upper back bones. The head bone is a child of the upper back bone in the model. I want to lower the shoulders in sync with the raising of the hips, but then the head follows and creates head jerking because the head follows the upper back. You can try to re-orient the
head, but it is very difficult to get right.
In a walk, there is some movement of the head but humans have a way of keeping their head steady despite moving other body parts. I hate to disconnect the head bone from the hierarchy if I can avoid it. I also tried putting a null in the model and constraining the head to it to see if it would override the bone chain, but it has not worked so far.
-
Attached is another update of the walk cycle project. This one changes the arm positions to be closer to the torso during the cycle as suggested by Mouseman.
I think this was a improvement. I also re-instated the delay in the arms coming forward at the start of the cycle that was originally suggested by Robcat and I
thought I had fixed, but as Mouseman pointed out, it needed to be more prominent. I also got the feet to be a little more stable. Definitely not perfect, but better than before.
Mouseman also thought the feet seemed to be rushing toward contact with the ground on the forward foot in the stride. I intentionally made it this way
because the Robcat walk tutorial clearly shows that the back foot stays back in contact with the ground longer than you would think. This causes a
slight off-balance position at a point a little past half way and the front foot has to compensate by coming down faster to maintain balance. I think the
relative speed of the walk has a role in this timing. If the walk is slow, the foot doesn't have to move so fast, but in a faster walk, it does have to move fast.
The female walk cycle I referenced in a previous post can be found by searching YouTube for Female Standard Walk. It gives a side and front view of a walk cycle,
including start and stop. It seems a little exaggerated, but it is still helpful.
Robcat caught the fact that I had to use a different hip position on the 2nd part of the walk for one of the key frames (what an eye!). For some reason, the model seems
symmetric, but I get knee popping on one side and not the other for the same position on the opposite side. I am not smart enough to figure out why, so I cheated and
just raised the hip by one key click. I did not think anyone would notice (I was wrong).
Once again, thanks to everyone for the suggestions.
Hit Man
in Work In Progress / Sweatbox
Posted
I am working on a simple animation of a hitman-type character. I have the poses worked out and before I go further, I would like to know if they read well enough to understand what is happening. If someone would take a minute to look at the attached animation and let me know if it is clear, I would appreciate it.
hitman2a.mov