Pepsi takes a tumble: Animation

After making the pepsi bottle by following the tutorial, i thought an animation would be cool… If one can of Pepsi is good, 1000 is better, right?

See for yourself:
High Quality Divxhttp://mpan3.homeip.net/personal/b.avi
High Compression Divxhttp://mpan3.homeip.net/personal/c.avi

The physics is recorded from the 2.40 Alpha 2 Game engine using the “record to ipo” feature. Took 5 minutes to step through the 500 frames. Rendering took around 100 hours in total with AO, area lamp and 8x motion blur.

Athlon 2.4Ghz
1.5GB RAM
Win XP :expressionless:

Attachments


LOL! That´s kind of funny - especially since you´re probably
mocking 3dstudio max´s ragdoll-demo :smiley:

Hmm, never seen the 3dsmax’s ragdoll demo, but i guess great minds think alike :wink:

i hate to walk into this room when thirsty.
-I take it this is a new feature. i havent gotten the chance to play with the beta yet. I dont ge tmuch time to play around with blender as much as i used to.

 great test, the physics look pretty impresive except at one point i saw a can gitter, maybe it was about to blow :D .

You are quite right. I’ve noticed a few cans do sudden rotations in mid-air for no apparent reason. Upon closer observation, it seems it’s an interpolation related problem. Somehow at certain angles, (89 to 91 for example), the IPO goes beserk and cause the can to do a fast rotation in midair.

I actually really enjoyed watching that. Very entertaining for some some reason. :smiley:

Though there is a lot of twitching going on, especially near the very end.

[quote=“mpan3”]Rendering took around 100 hours in total with AO, area lamp and 8x motion blur.
quote]

looks good - but 100 hours? you got more patience than I’ll ever have! :slight_smile:

I just let it render from the command line at lowest priority. My PC is “always on” so it doesn’t hurt for it to do something useful at midnight. :wink:

Gaaar, now you got me all thirsty :expressionless:

Great composition nevertheless! Impressive to say the least!

WOW! That is great!

I am going to have to try that “physics” sometime. I am glad to see that the Pepsi Tut cam in handy.

Just a friendly reminder; the Blender Sumo physics engine still sucks (I am sorry but it’s true).

  1. It can’t handle high polygon meshes well, even if they are inside a “sphere” collision box. prepare for some extreme slowdown.
  2. too many objects is also bad for performance. The above movie is baked @ 0.1 FPS on a ‘FAST’ system.

An alternative to the GE -> IPO bake is this:
http://mpan3.homeip.net/personal/Collision.zip

Amazing! Lots of glitches with the motion, but it still looks great.

I’ve decided to share the .BLend file… A hefty 23MB packed and COMPRESSED w/ 2.40 Alpha 2

http://mpan3.homeip.net/files/Blender_PepsiSim.blend

Sorry about the slow server, I’d recommand a segmented dl accelerator.

(most of the file is IPO data, before baking the .blend file is less than 2 MB)

Well i’ll mirror them for yeah :smiley:

The .blend
http://www.h4x.snipanet.com/blender/pepsi_takes_a_tumble/Blender_PepsiSim.blend

High Quality DivX
http://www.h4x.snipanet.com/blender/pepsi_takes_a_tumble/b.avi

High Compression DivX
http://www.h4x.snipanet.com/blender/pepsi_takes_a_tumble/c.avi

that animation was wonderful!
i would love to see a tut or some explanation
of how this physics works. i would love to do some
animations along this line.

SUPER JOB!

great simulation and composition camera running :smiley:

Great animation, very well done, I like the camera movement.

Thanks! Looking at all the replies really justify the hours spent rendering this beast…

/Me wants Dual Opteron … :x

That’s awsome! :slight_smile:

The cans that went directly in front of the camera and made it flicker black were annoying. I’d suggest putting an invisible collision mesh parented to the camera so that they would bounc off instead. Really nice though! :slight_smile:

really nice job.

to speed up the physics simulation and baking to IPOs, use a low res proxy cylinder. assuming you’ve duplicated your cans using alt d, you should be able to relink the mesh datablock for one can to the hi res mesh and the rest will follow suit.

convex polytope is a better bounding box to use, and you mentioned you used sumo as the physics engine - i think it would be better to use bullet.

let me know your thoughts on this - whether i am off base or not. i only started exploring the physics engine a couple of days ago so i could be wrong.