IRTC anim topic: Impostor!

General discussion about JPatch

Postby dcuny » Fri Apr 15, 2005 11:22 pm

I love it! :D

Great work, get some sleep - you deserve it.

You'll need too... With the animation out of the way, I'll start asking for new features again! ;)
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby dcuny » Sat Apr 16, 2005 8:57 am

Hrm... Seeing that there only appear to be three submissions for this round (so far, anyway), there's a good chance that you might make the top three this time around. ;)
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby sascha » Sat Apr 16, 2005 10:19 am

Yes, if nothing goes wrong, a top three position should be within reach :)
But seriously, I hope that there are some panel judges, so they don't have to cancel the voting (three judges are not a hell of a lot).

With the animation out of the way, I'll start asking for new features again!

:)
That reminds me: I've posted a few suggestions regarding JLipSync here some days ago. I thought you maybe missed it because of the cookie problems.

Back to JPatch:

When working with it, I discovered a lot of annoying things: The modeler is quite usable, but there are some bugs with morphs. You have to be very careful when deleting a morph - it seems that it gets confused when the morph slider is not at it's 0 position. You can even delete a morph while editing it, putting JPatch over the edge :? I think, when adding a new morph, it should go into "edit morph" mode immediately.There are some problems with keyboard focus as well.

For the animator, the most annoying thing was that there is only one camera, closely followed by that fact that everything you do only affects the current keyframe. If you are at frame 100 and change a lightsource's intensity, it results in a change of intensity (e.g. between frame 0 and 100) - if this is not what you want, chances are you discover it after rendering. I don't know how many shots I had to rerender because of things like that. If you change the lenght of the sequence later, thigs get even worse - if there are keyframes outside the range, you can't see them, but they are still affecting the animation...
Next, the animator simply enumerates all morphs. I chose this approach because it may be possible to have two morphs with the same name. A problem is - if you delete a morph, the remaining morphs will be mapped to the wrong motion-control-curves when the animation is loaded. I think I'll have to check for the morph names as well - if the name does not match, it would ask the user for help. Another way would be to assing a unique ID to each morph (e.g. generated with a timestamp and a random number).
Then, there should be a way to select multiple motion-keys and change their position in time (move them left or right), cut'n'paste, etc. Hacking in the xml file is not really user friendly.

I've used u/v mapping twice in this animation. One time for the floor morph (planar mapping) and one time for the tomato soup logo on the tin can (cylindrical mapping). It was hard-coded - but I think I can quickly implement it as a feature the following way:
In the material definition, you can choose between some u/v modes: off, planar, cylindrical, spherical, map.
planar would require to enter a vector (the normal to the plane), cylindrical too (the cylinder axis), map would be a custom u/v map (to be implemented in the future).
A feature to render just a subset of the sequence (but with the images saved with proper frame numbers) would be helpful too.
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Sat Apr 16, 2005 11:30 am

Sascha wrote: I've posted a few suggestions regarding JLipSync here some days ago. I thought you maybe missed it because of the cookie problems.
Yes, I missed it completely. I'll answer those on that thread.

When working with it, I discovered a lot of annoying things...
I know the feeling. :D

If you are at frame 100 and change a lightsource's intensity, it results in a change of intensity (e.g. between frame 0 and 100) - if this is not what you want, chances are you discover it after rendering.
I remember hitting on this a while back when we were dicussion how motion curves should behave in the first place.

I had suggested that perhaps for the first key, JPatch could insert a prior key at zero, but that's not always the right behavior, either.

Probably the best solution would be to have a fast preview renderer. (I've been working on the zbuffer code, but I'm stuck at the moment on how it interpolates the z value).

Hacking in the xml file is not really user friendly.
I think the main advantage of XML is that you don't have to deal with endian issues and such, but I don't really think of it as a format I'd actually want to work with at a low lever. At some point, the XML file is just going to be too complex to work with.

planar would require to enter a vector (the normal to the plane), cylindrical too (the cylinder axis), map would be a custom u/v map (to be implemented in the future).
Or you could just use the default normals, since JPatch assigns a normal (whether you want one or not).

A feature to render just a subset of the sequence (but with the images saved with proper frame numbers) would be helpful too.
I think a couple times re-rendering scenes to fix lighting issues will probably convince you that being able to render to seperate channels (diffuse, specular, etc.) is a real time saver, too. I think this is one of the next items for A:M.
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

HDVT version

Postby sascha » Mon Apr 18, 2005 11:50 am

I plan to render it again at HDTV 720p resolution (1280x720). If you discover any errors or things that could be quickly improved please let me know (so I can correct it before starting to render it :) )

Here's a still.
Here's here one at 1080p resolution (1920x1080).

I'll have to use higher resolution image-maps and probably work on the textures and add some more details to the objects.

There are two divX trailers in HDTV here:
http://www.divx.com/hd/
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby animohn » Mon Apr 18, 2005 1:41 pm

I hope you render the HD version without the subtitles. It looks so much better without them.
animohn
 
Posts: 31
Joined: Mon Jan 31, 2005 4:19 pm

Postby dcuny » Mon Apr 18, 2005 10:08 pm

Well, I've seen the other two IRTC submissions.

Chess has a lot of nice technical bits - focal blur, camera moves, cool attacks... But the ending left me rather puzzled. :?

Crime and Punishment was rather disappointing. There are some clever bits (such as the sign reading "Ask us about our new anti-fraud program"), but it sort of fizzled in the end... Too bad, I'm a big Rusty fan. :(

Personally, I'm hoping you make the #1 spot (third times the charm!), but I'm a bit biased. :D
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby sascha » Tue Apr 19, 2005 8:08 am

Chess has a lot of nice technical bits - focal blur, camera moves, cool attacks... But the ending left me rather puzzled.

At first I did not see the relationship to the topic. It took me several times to watch it until I noticed that the pawn at the end changes its color before attacking the king.
All in all a very nice animation, the chess pieces are well modeled, and I like their hands and weapons "morphing out". My only complain is the use of depth of field in the last shot. Changing the camera focus from the king to the pawn would help to drag the attention to the pawn (that changes its color). But the camera first focuses on the king, then on infinity and then on the pawn, which is quite misleading and makes the scene difficult to understand.
Some physics simulations of the falling pieces would have been nice too (e.g. have them rebound instead of sink in the floor).

Crime and Punishment was rather disappointing. There are some clever bits (such as the sign reading "Ask us about our new anti-fraud program"), but it sort of fizzled in the end... Too bad, I'm a big Rusty fan.

I did not read the sign until you mentioned it, but I like how the ad display works. I also like the "holographic" projection of the letters a lot.
Voting will be difficult, I don't know yet which one will get more points from me.
But I think that only two voters (for each entry) is too little. I fear that if they don't have some panel judges, they'd have to cancel voting :(
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Tue Apr 19, 2005 10:08 am

sascha wrote:
At first I did not see the relationship to the topic. It took me several times to watch it until I noticed that the pawn at the end changes its color before attacking the king.
I got that right away - I just didn't "get" why it happened. All the prior animation focused on "normal" chess, so to have one of the pieces change at the end - out of the blue, with no explanation why - was puzzling. I watched it a couple more times to see if I had missed some clue, but if I had, I didn't see it.

Some physics simulations of the falling pieces would have been nice too (e.g. have them rebound instead of sink in the floor).
The "explosions" on the ground didn't quite work, either.

Still, it's got some very nice visuals.

I did not read the sign until you mentioned it, but I like how the ad display works. I also like the "holographic" projection of the letters a lot.
Yes, I noticed those innovations as well.

I had a bit of trouble reading "Rusty's" display this time around - it seemed to go faster than usual.

But I think that only two voters (for each entry) is too little. I fear that if they don't have some panel judges, they'd have to cancel voting :(
That would be too bad. :(
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby J. Baker » Thu Apr 28, 2005 4:51 am

Nice work on "Imposter", looks great! JPatch has come a long way from when I first came across it. I'm going to have to start modeling soon. :D

To dcuny: Nice lipsync program. Getting ready to write my own for KoolMoves.
AMD Semprom 2400+ - PQI Turbo 512MB DDR PC2700 - ATI 9100 64MB DDR PCI - Samsung SP0411N - PcChips 848ALU 2.1 - WIN XP Home SP1
J. Baker
 
Posts: 13
Joined: Sat Jun 12, 2004 6:06 am
Location: USA

Postby dcuny » Thu Apr 28, 2005 5:13 am

Be sure to check out Papagayo from Lost Marble - it's a nice clone of Animation:Master's dope sheet. Now that I've played with it, I've got to rewrite JLipSync... :?
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

Postby J. Baker » Thu Apr 28, 2005 6:25 am

dcuny wrote:Be sure to check out Papagayo from Lost Marble - it's a nice clone of Animation:Master's dope sheet. Now that I've played with it, I've got to rewrite JLipSync... :?


I saw that, very nice indeed. :D Keep up the good work with your program.
AMD Semprom 2400+ - PQI Turbo 512MB DDR PC2700 - ATI 9100 64MB DDR PCI - Samsung SP0411N - PcChips 848ALU 2.1 - WIN XP Home SP1
J. Baker
 
Posts: 13
Joined: Sat Jun 12, 2004 6:06 am
Location: USA

Postby J. Baker » Sat May 14, 2005 4:22 pm

Just wondering, how long did it take to render the imposter and on what system?
AMD Semprom 2400+ - PQI Turbo 512MB DDR PC2700 - ATI 9100 64MB DDR PCI - Samsung SP0411N - PcChips 848ALU 2.1 - WIN XP Home SP1
J. Baker
 
Posts: 13
Joined: Sat Jun 12, 2004 6:06 am
Location: USA

Postby sascha » Sat May 14, 2005 6:33 pm

Just wondering, how long did it take to render the imposter and on what system?

I have a Pentium IV; 2.8 GHz; 1GB Ram; Fedora Core 2 Linux.
I used a quite high setting for antialiasing, but aside from that there's nothing special (ok, most surfaces are reflective and there a a number of lightsources, but that doesn't really stress POV-Ray. If I'd had more time, I'd have used focal blur and soft shadows in some shots thought). A typical frame took about 30 to 90 seconds, including the time to generate, parse and render it.
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

Postby dcuny » Mon May 23, 2005 6:20 pm

Congratulations! Imposter ranked number one! :D

As an added bonus, you were top in all categories as well - technical, creative and artistic.

Well done, you deserve it!
dcuny
 
Posts: 2902
Joined: Fri May 21, 2004 6:07 am

PreviousNext

Return to General discussion

Who is online

Users browsing this forum: No registered users and 1 guest

cron