HOME | DD
Trackdancer β MMD Tutorial on Chibi-sizing Motion Data
by-nc-nd
Published: 2012-08-05 09:17:07 +0000 UTC; Views: 37483; Favourites: 671; Downloads: 428
Redirect to original
Description
Simple method on how to scale down motion data for use with chibi-sized models.Important Note:
The figures I used are for demonstration purposes only, they cannot and should not be applied in all circumstances. You will need to make adjustments to match each individual model and motion data set. So for example, when I did my Chibi Music Class video, although all 5 chibis used the same motion, each one required a different set of adjustment ratios (well 4 of them did anyway, Chibi Zatsune Miku and Chibi Miku used the same ratios).
Additional Notes: courtesy of Kas97
1/. Nendoroids will need to use ratios of around 0.4:0.9 - see his her notes in the comments below for specific details
2/. A useful tool (software) for fine tuning motion data to fit a specific MMD model: http: //yumin3123.at.webry.info/200811/article_1.html
Example video links (YouTube):
MMD Chibi Music Class - youtu.be/OYLAGXwKzaI
Related content
Comments: 224
π: 0 β©: 0
π: 0 β©: 1
π: 0 β©: 0
Seriously, this literally saved all my chibi models from moving awkwardlyΒ Β Thank you so much.Β
π: 0 β©: 1
So is that the same with camera? For example if I wanted to used a motion on a chibi/child/smaller model and the motion comes with a camera motion as well?
π: 0 β©: 1
Camera motions don't scale down well. Best to either manually adjust them or make these from scratch.
π: 0 β©: 1
I see thank you π
π: 0 β©: 0
There is some tutorial of this but for high models??
π: 0 β©: 1
Principle explained here is the same for larger models.
π: 0 β©: 0
In case of group dances where models move around and change positions with each other. How could I go about that? or it it's possible? I tried this method and it doesn't work on those motions.
π: 0 β©: 1
Well it is actually possible but requires a reasonably good understanding of how to scale motions. The trick here generally is to figure out the correct adjustment for the start location of each performer.
If you understand fully what I am saying here, scaling group dance motions is just a basic math exercise and some trial and error adjustments. If you don't understand the process fully, then it will be really tough to do.
π: 0 β©: 1
Hmm. You see I'm trying to make my Animal crossing models that I ripped as compatible with other motions as possible. They have big heads and little legs and that certainly makes things harder. When I scale them they lose movement but if I scale them less their legs are spread out. By losing movement even if I use the master bone to move them to right or left they still hit each other or maybe I didn't understand what you were trying to tell me. Anyway I want to understand no matter what so please be patient with me.
π: 0 β©: 1
MMD motion data are a record specific positions.
So for example if the data set states that the legs are moved from position X1,Y1,Z1 to position X1,Y1,Z1, when loaded on to another model of the same proportions you will get the desired expected results.Β
When this same data is loaded on to a chibi version of an adult sized model, the data will also move the leg from X1,Y1,Z1 to position X1,y1,Z1, but of course this doesn't give the desired results and the chibi will stretch unnaturally to try to reach the recorded end position but can't reach it as the chibi's legs are too short.Β
So when downsize a motion you adjust the motion proportionately using a ratio that reflects the size of the chibi relative to its adult sized equivalent. For example, Chibi IA is roughly 50% smaller than her adult IA form. So a motion that works well on IA will stretch Chibi IA unnaturally. Downsizing the motion by 50% then gives the desired results and this principle is shown in the tutorial above.
When you want to downsize a group dance the principles still applies but it becomes a two step process. Specifically:
1. Downsize the motion to suit the chibis - they need to be approximately the same size for best results so that the motion can be downsized using the same ratio.
2. Downsizing the motion does not account for the start locations and these parameters remain as originally set and so the close coordination between the performers are lost. To correct this, the general principle is to downsize the start locations manually as well using the same ratio. But even here, some degree of manual fine tuning to avoid collisions is almost mandatory.
NONE of these principles applies to heavily deformed models.
Everything discussed here works for "normal" chibis which are more or less a smaller sized version of their adult equivalents and have "normal" human proportions.
When you have a heavily deformed model in terms of their bone structure: big heads, short limbs; none of these methods work because the original motion was designed to work with normally proportioned bone structures.
Why?
Well simply when your model has short legs relative to the torso (for example) when the motion mandates that the legs be moved fromΒ position X1,Y1,Z1 to position X1,Y1,Z1 you're going to get bizarre results because the motion just wasn't made to accommodate aΒ model with distorted proportions.
So is there a solution?
Actually, yes. This is to adjust the motion manually. Specifically:
1. Downsize the motion to best match the proportions of the torso and legs. You don't really need to account for the head so much as heads don't move around that much relative to the limbs.Β
2. Then manually correct the motion as applied to the distorted model. So if the legs still stretch out too much, you'll need to correct this manually.Β
Yes it is a lot of work, but then this is the price you will have to pay if you're applying data made for a normally proportioned model to a model with a heavily deformed humanoid structure. So if you're looking for a "quick fix", there is any.
π: 0 β©: 1
Ok thank you so much! I just wanted to know if there was a quick way out of it but I guess not XD
π: 0 β©: 0
Thank you for this, but I'm in a bit of a situation. The model is a tad bigger and more cartoony in shape. Her head is bigger and the main problem is that her arms go into her body. How should I approach this?
π: 0 β©: 0
Thank you for this, it's absolutely wonderful! I do have a question though, do you have any tips on figuring out the ratios? I have a model whose proportions are rather different (I can send you an image if you'd like to see what I mean). I would love to be able to make her dance, but I just can't figure out the proper ratios!
π: 0 β©: 1
Treat the default Miku as 100% and scale by eye from there.
There is a more scientific way of doing it if you want to do the maths, but it can be a real pain just measuring the models. There's also a program that does it but it only works for PMD models plus I don't think it's available anymore. Besides, usually, the motion makers don't tell you which model the motion was originally made on and this is a big factor in the equation.
π: 0 β©: 1
Okay then, thank you. I've managed to get it working....better. Her legs are still doing strange things such as her knees are bending backwards....but at least it doesn't look like someone is grabbing her by the legs and trying to tear her apart. Lol.
π: 0 β©: 1
Her legs are bending backwards because the leg IK chain is incorrectly set up. There's a tutorial in my gallery explaining how to fix that issue.
π: 0 β©: 1
I found it, and it's fixed. Thank you! Your tutorials are wonderful.
π: 0 β©: 1
Hehe, they are aren't they? ^^
Glad to be of service.
π: 0 β©: 0
I have a problem. I have a model which is a little smalleer than the others, and when I load the motion data in that model, the facials of the motion don't work...Do you know what I can do? I tried doing what you explain in the tutorial, but It didn't wor for the facial expression D;
π: 0 β©: 1
Facials functions like programming subroutines. They only work when called/triggered. If your model has non-standard facial labels, they won't work with motions that use regular facial labels.
So for example the motion has a call for a facial labeled "BLINK", and your model has a blink but it's labeled something else, then there's no way that motion will make your model blink.
π: 0 β©: 1
never mind, I actually fixed it > u < but thank you so much for the help! <3
π: 0 β©: 0
*I need a tutorial on how to make the camera perfect for the small MMD models.
*I've download motions and then load it onto the small model and then load the camera, but when I play the motions, I can't see the small model's face because of the camera so I really need help.
*Those cameras only fits the adult-sized models. I cried.
π: 0 β©: 2
You can adjust the camera motion data to suit the model. The camera's motion is scalable. Also it's sometimes easier or more effective to create your own camera motion.
π: 0 β©: 1
Idk how to adjust it...
π: 0 β©: 1
There is a similar panel under the main menu across the top that allows you to adjust the camera angle. I don't mention it as when I did the chibi motion tutorial this feature wasn't available. But it is there, it does work and I am not awfully impressed with it's function. Much better to make your own camera motion for these kind of situations.
π: 0 β©: 1
simple just to adjust the camera with the height and angle
π: 0 β©: 1
Ahh, what an incredible help this knowledge will be! Thank you so much!
π: 0 β©: 1
Happy to hear that you've found this useful.
π: 0 β©: 0
Link of Motion conventer don't work. Do you have this program? do make a note for me?
π: 0 β©: 1
I didn't keep my copy of the program as I've hardly ever needed to use it plus it only works with PMD models.
π: 0 β©: 0
DO You know how to put the models on a separated side cause every time I put the same motion on a model and another model for them to dance together They get on the same spot Do You Know How To Fix It?
π: 0 β©: 1
Relocate one of the models using it's mother bone.
π: 0 β©: 0
Why does it not work for me? Why I do it the models go weird again and when I click back on Multiply of frame it still be at 0.1 instead of what I set it.
π: 0 β©: 1
Not sure what you're doing to cause what you report to be happening. But I can tell you that you can only set these values once in general, so if it messes up, you need to redo the entire set up from scratch.
π: 0 β©: 1
Ohhh thank you! On the same night I have fixed it but thank you for the reply! n.n
π: 0 β©: 0
Β I dl'd a model and it said "1.put all the texture and model data in the same folder 2.launch mmd and go to "import data" and select the folder then load" how do I do this? plss help me I'm a total noob to mmdΒ
π: 0 β©: 1
1. The model data and the texture files all go into the same folder.
2. Load the model as usual. Easiest way to do this is to just drag the model data file's (usually a .pmd or .pmx file) icon from its folder directly into the MMD user interface.
π: 0 β©: 1
thank you for the helpΒ
sorry to ask again but I dont understand at all I only understand no.1 I'm so sorry I'm bad at english XC
π: 0 β©: 0
can you do a tutorial on the vmdconverter?
π: 0 β©: 1
Afraid not as I've never used it.
π: 0 β©: 0
thanks for this tutorial
It's really useful
π: 0 β©: 1
| Next =>