IRTC entry

General discussion about JPatch

IRTC entry

Postby sascha » Thu Nov 25, 2004 3:37 pm

I've tried to model a rock. What do you think?

Image

I'm not quite happy with it, as it doesn't look funny :?

Here's the JPatch file: rock_4.jpt
Feel free to change it!
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Thu Nov 25, 2004 6:02 pm

Cool! They look sort of like the rocks at Easter Island. Except, of course, that those guys actually had legs, they were just buried in the ground. :D

The main thing that should probably be changed is to make the teeth parallel, so they match up to phonemes a bit better.

That reminds me - I saw some TikiMon figures the other day:

ImageImage

Nice texture... POV-Ray?
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby sascha » Thu Nov 25, 2004 8:36 pm

They look sort of like the rocks at Easter Island

Yes, I had the same thought (although that wasn't intended)...
Image
But that's about the size of the rocks I had in mind - for smaller rocks the "In a million years, we'll still be here" sentence sounds a bit unbelievable.

You're right about the teeth, and by looking at your TikiMon images I thought the eyes needs to be larger. I'll try to improve it.

Nice texture... POV-Ray?


Yes, but it's very simple - just a marble pigment with some turbulence and a granite normal (the texture is included in the .jpt file). My "InyoPatterns" can do that (and better) easily.
The main problem with povray is that it doesn't support referece geometry, so the texture can't be animated.

Once I've got the new triangle export code (including reference geometry) up and running I'll try to get reference geometry support and the patterns working in Inyo...
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby sascha » Sat Nov 27, 2004 10:29 am

Here's an improved version, this time rendered with Inyo...
Image

I have a question about ambient occlusion: How is it affected by the ambient and diffuse value of the material? Does it only use ambient (is it gone when ambient is 0 and strongest when ambient is 1)?
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Sun Nov 28, 2004 12:35 am

No, it just skips the test if the material is not diffuse. It's there to prevent the code from running on materials that are transparent, although it's probably not the best test. There's a boolean flag isTransparent that's probably more appropriate to use.

It scales the power of the illumination by the world.skyPower, which is the strength of the skylight. Since ambient occlusion is supposed to simulate scattered skylight, I figured it was as good a value to use as any.

Typically the value is additive. If you choose to subtract the amount, sets the initial value of the light contribution to 1.0, and then subtracts from that for each AO sample that occludes.

I should point out that when oversampling is at 3, there are 9 samples per pixel. Each of these samples sends out 16 AO samples, which makes 144 AO samples per pixel. That's probably a bit higher than necessary. :?

I see teeth have been repaired!
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby miyoken » Sun Nov 28, 2004 12:46 am

Hi.
I've tried to model a rock. What do you think?


It's nice.

The second rock has a somewhat fearful feeling.

"funny" is difficuft...
Image
miyoken
 
Posts: 39
Joined: Mon Jun 07, 2004 11:16 am
Location: Japan

Postby sascha » Sun Nov 28, 2004 10:39 am

Hi miyoken

The second rock has a somewhat fearful feeling.

You're right, It somehow reminds me of a totem pole now :?

I've tried to modify it according to your suggestions. The new model can be downloaded from here: http://www.jpatch.com/temp/rock_7.jpt

Image
I'm not an expirenced modeler at all, so if you can make it look less fearsome and more funny, please go ahead!

David, thanks for the hint with the AO samples - I wandered why it was so slow :oops:
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Sun Nov 28, 2004 8:26 pm

I think aiming for something more like the Easter Island statues would be a good thing. That means making them a giant head placed on a tiny body. As they currently are, they look more like faces carved onto round stones - the face isn't distinct from rest of the body.

Notice also that a large, flat chin seems to be common among all these figures. The noses are carved sqare, not rounded. Even with considerable weathering, the squareness of the nose remains. The same goes with the jaw - the general shape is very "square", not rounded.

Still, there's nothing wrong with your design. It's clear what the statues are, and they had a style of their own. That's a good thing.

But I think you should make the eyes white instead of stone, as well as the teeth, like the TikiMon figures.

Feh... now that I've criticized it, I'll need to have a go at it myself. :?
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby dcuny » Mon Nov 29, 2004 7:25 am

I started working on a simple model (moai.jpt):

Image

One problem is that there's a hole above the nose in the forehead. Any idea why I can't fill it?
Last edited by dcuny on Mon Nov 29, 2004 10:21 pm, edited 1 time in total.
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby sascha » Mon Nov 29, 2004 12:47 pm

Yes, because all the points are on the same "spline"...
JPatch is not good for polygon-based modeling!!
You could fill the hole by adding a line (and dividing it into two triangles).
It would be easier to start from an outline curve and extrude or lathe it... stay in "round" mode to better see how the splines are connected and use the "peak" mode only at the end, to create sharp edges where desired...

The many orange squares indicate that there is a problem with this model - selecting everything and switching to "round" mode shows that the patches can't be connected smoothly - this is because the splines are connected in an odd way...
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Mon Nov 29, 2004 7:18 pm

OK, I'll have to keep that in mind.

Is there any way of attaching an error message to a particular point, so I can know why it's orange? If they all indicate the same error, there's not really any point, though.

This "one the same spline" business is a bit tricky for me, because it's not visually obvious which splines are the same... :?

That sort of reminds me of the "problem" I have when attaching splines. I'll want something like this:

Image

I'll start out by attaching a spline to the end of an existing spline's end:

Image

And then extend it, but I'll get this:

Image

This is obviously nothing close to what I want! The solution I currently use is to create a new spline:

Image

and then attach it to the existing spline:

Image

One of the problems is that you'd need to be able to tell which spline you want to attach to (if any). I can't think of any way of doing this that's less complicated than just welding a new spline's centerpoint like I'm currently doing.
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby Guest » Mon Nov 29, 2004 9:01 pm

The orange square doesn't decessarily mean that there is an error. Here's how it works:
Take a single spline - all controlpoint are shown as yellow diamons - meaning that there is no other point "attached" to them. When you weld another point or spline to a start of the curve, it is "appended" - there's still just one spline, so all points are still shown as yellow diamonds.

What happens when you weld another point or spline to a point in the middle of the spline is different - I called it "attaching". Two splines meet, and the intersection points are welded - they are still individual points (on different splines) with different tangent vectors, but share the same position.

When patches are connected, usually two splines meet at the corner of a patch - so a patch corner are two attached control-points. This is the normal situation, and such points are displayed as tiny red squares.

If a third (or 4th,...) controlpoint is attached to such a point, it is shown as (larger) orange square. There are situations where this is desired - but it can also happen when a points is welded to the wrong spline by accident. So if you see an orange square and don't know wyh it's orange, it's possibly because of a mistake, and it should be examined.

A "red" controlpoint normally is a shared corner of four patches (except on boundaries). Four patches will, in general, connect smoothly. An orange controlpoint can indicate that more than four patches meet at that corner, and patch continuity cannot be guaranteed - this usually results in visible creases in the surface.
A special case (which is treated seperately by JPatch) is the center of a lathed object (e.g. a cone) - if the lathe was done with 8 segments, 4 splines will meet at the center point. As long as this point is not moved out of position after the lathe operation, JPatch will ensure smoot continuity, but if you move the point creases will appear.

Image

Here are some exapes: Left: the standard case. Second from left: a lathed object with an orange point in the center. Third from left: Four patches share a common edge - if this is desired, it is ok - two pairs of patches will connect smoothly. Right: This is a problem: Six patches meet at the point marked orange - JPatch cannot connect the patches smoothly, this results in visible creases in the surface. All of these problematic topologies can be resolved to "red" controlpoints using either hooks of five-point patches.

Most orange controlpoints that actually indicate a problem are caused by improperly used three-point patches.

See also "how to avoid creases" in the documentation.

Image
This is another problem. JPatch connects the new point (or spline) to the first spline it finds. If it doesn't work as desired, try to connect the point the other way round (in the case of your images, first connect the spline that extends to the right (and appends to the horizontal curve), then attach the vertical spline to it (just as you did). There is currently no other way to do this.

If you'd like to create a sharp corner (e.g. a corner on a grid of splines), and you try to weld the two splines meeting at the corner, JPatch will automatically append one spline to the other. In this case, this is not what you want. The workaround is: Extend one of the splines by one more point. Then weld the other spline end to the disired point. After that you can delete the previously added (and unneeded) point.

It's a bit complicated, I know, but with a little practice you'll quickly know what JPatch will do, and you'll use these "workarounds" virtually automatically. If you know a way to make this more user-friendly, please let me know, I'll happily implement it.
But I think even Animation:Master has the same sort of limitations (that can be worked around using the same procedures).

This "one the same spline" business is a bit tricky for me, because it's not visually obvious which splines are the same...

Imagine a single spline, made of 4 controlpoints that is closed. It represents a loop with 4 points (and would form a patch), but there's a rule that checks that condition and will prevent it from making a patch. Imagine a lathed object lathed with 4 segments - there would be a lot of invalid patches if these rule wouldn't exist.
If the tangent mode is set to "round" it is usually obvious which points are on the same spline. In cases where it isn't obvious, you can use the "extend selections" feature (ENTER key) for curves: Select a curve segment (TABULATOR) and hit ENTER - this will select (and highlight) the entire spline.

I hope this explaination is of help. I remember my first steps with sPatch... I ended up with a model full of creases, and no clue about what was happening. But with a little practice, you'll quickly understand how it works and why it does certain things. At least JPatch offers 5-point patch hand hooks support, which can be used to resolve situations that always lead to creases in sPatch.

There are some sPatch and A:M tutorials that cover these subjects. Just let me know if you have more questions about it.
Guest
 

Postby dcuny » Mon Nov 29, 2004 10:18 pm

Thanks for the detailed explanation. :D

I had forgotten about the Extend Selection option. I'll have to play with it some more.

As you mention, these are issues inherent with using patches, and even A:M has the same sort of issues. If you ever get around to supporting subdivision surfaces, I'm sure they'll have their own 'gotchas'.

Speaking of 'gotchas', one thing I keep tripping on is the use of the Auto Mirror tool without first using the Align Axis tool. The result looks like it's welded together, but it's not. Perhaps there could be a warning in these cases?

Thanks again!
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby dcuny » Mon Dec 06, 2004 11:57 am

Since my modeling skills aren't up to the task, here's an image I put together in The Gimp:

Image

Yes, I know: my four year old daughter could draw better than that.

Those blobby things in the foreground are supposed to be sheep. Since they don't do much other than munch grass, I figured they might be fun to have in the scene. Plus, you could add the sound of sheep to the soundtrack.

I was thinking of something along the lines of Shaun from Aardman's "A Close Shave."

I was hoping to tape the vocals tonight, but couldn't find the microphone. The darned thing's always in the way when i don't want it; now it's gone missing. I'll probably have to borrow one from my brother. :?

The good news is that I got the computer desk mostly cleaned off. :D
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby sascha » Mon Dec 06, 2004 6:35 pm

Hehe, I used the Gimp for some simple things (taking screenshots for example), but never understood its layer conecept. I'll have to play with it some day...

A sheep would be nice (if time allows). It would need a fur texture (that's gonna be tricky) - and we'd need grass...

I'll add morph targets to the rock soon - can you make a JLipSync track when the voice-recording is done?

I'm not sure about the camera. Will it be only a single shot, or do we switch between closeups of the two rocks?

I also thought about adding water (a lake with the rocks reflecting in it) just to show that we've got a raytracer 8)
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Next

Return to General discussion

Who is online

Users browsing this forum: No registered users and 1 guest

cron