sprockets TV Commercial by Matt Campbell Greeting of Christmas Past by Gerry Mooney and Holmes Bryant! Learn to keyframe animate chains of bones. Gerald's 2024 Advent Calendar! The Snowman is coming! Realistic head model by Dan Skelton Vintage character and mo-cap animation by Joe Williamsen
sprockets
Recent Posts | Unread Content
Jump to content
Hash, Inc. - Animation:Master

Recommended Posts

Posted

This is my latest WIP. I'll post the actual render when I'm satisfied with the lighting and details, but thought I'd share this version with a PS2 Neon filter applied. It's currently my Desktop.

599__9G.jpg

  • Replies 37
  • Created
  • Last Reply

Top Posters In This Topic

Posted

Hi guys:

 

Here's what I've got so far. I'm trying to go as realistic as I can, but it's pretty tough for me. The post work lens flares don't help either. It's fun to tinker though.

 

Eric

599__10.jpg

  • Hash Fellow
Posted

That looks wonderful!

 

 

One thing about the lens flare... the various flares should have a common point of convergence (usually the center of the photo). Unless there's a reason they don't?

Posted

Thank's for the encouragement gents. It's amazing how many variables there are to getting that render just right.

 

Rob, the lens flares are more or less an artistic touch and were more or less just thrown in, sorta like the bling on a pile of treasure. Is it even possible to have multiple flares on one picture? So much for realistic :lol:

  • Hash Fellow
Posted
Is it even possible to have multiple flares on one picture?

 

Yes, any sufficiently bright point of light can cause a lens flare. They will all radiate around the axis of the lens.

Posted
Love that first image ,that looks better than the real thing ,excellant

 

Was it really rendered with 49 passes? For 7x7 Passes it has a bit much of aliasing at the outsides...

Other than that: Very well done! Looks great!

 

*Fuchur*

Posted

Fuchur:

 

Good eyes, it was not rendered with those specifications. The frame with the title and data was something I made in Photoshop and belonged to one of the original renders. I did quite a few as I was modeling the car. There is also a problem with my final output as far as renderign goes. The image I posted is actually a screencap of the model after about 80% of the rendering was done. I did this because for whatever reason, I believe it has something to do with my ATI video card, if I let the render finish it will alias very badly. To show you what I mean, I'll start a render right now with 49 passes and let it finish all the way. Once done, I'll post it and you'll see.

Posted

Here is the render done to completion at 49 passes with 35% AO. Notice how badly it is aliased. I'm pretty sure it has something to do with my ATI card.

16A0.jpg

  • Hash Fellow
Posted
I'm pretty sure it has something to do with my ATI card.

 

the graphics card wont' play any role in anti-aliasing in a final render.

 

But that looks great! I'm not bothered by any aliasing.

 

Just to try, render it at 2x the res, do a 1 pixel blur, then shrink it down to this size. That ought to fix anything.

Posted

Beautiful model, as always, Eric, but... 49 passes... and aliasing around the lighter areas like that!!!? :o

What does it look like with 1 pass, for comparison?

I very rarely use more than three passes, because I can't stand waiting that long, and if 49 passes doesn't eliminate aliasing then what is the point?

  • Hash Fellow
Posted
Here's a render with 5 passes. The aliasing is even worse. I'm kinda at a loss here.

 

5 passes is expected to be worse than 49. What am I missing?

Posted

At this resolution I prefer A for it's clean sharpness, though the aliasing on the grill is rather obvious. B just looks like it has had a little gaussian blur applied. You could get the same effect much faster by bluring the frames in PS. I thought multipass was supposed to have some intelligence to it, focusing more on highlights and edges but this just looks like the whole image has been blured, at the cost of much higher render times.

Posted

First rate craftsmanship Eric!

 

May I please see a close-up wireframe of one of the wheel spokes showing the transitions into the hub and rim? I'm resplining my locomotive wheels and I'd appreciate seeing your solution.

  • Hash Fellow
Posted
At this resolution I prefer A for it's clean sharpness, though the aliasing on the grill is rather obvious. B just looks like it has had a little gaussian blur applied. You could get the same effect much faster by bluring the frames in PS.

 

A is Eric's original 49-pass render resized in Photoshop to 50%

 

B is the same render with a 1-pixel blur, then resized in Photoshop to 50%

 

I think B pretty much eliminates AA issues.

 

 

Another 3D app I have has some options on how anti-aliasing is done. It's basically the trade-off you see above. You can have sharpness or you can have smoothness, but you can't have both. They are opposite uses of the available pixels in the image.

 

 

 

 

 

 

I thought multipass was supposed to have some intelligence to it, focusing more on highlights and edges but this just looks like the whole image has been blured, at the cost of much higher render times.

 

Multi-pass is just a primitive brute-force approach to AA and motion blur. The regular renderer is the one uses some intelligence to pick out things that need effort put into them. Perhaps it can be modified to offer the option of a 3rd anti-aliasing stage on top of the two it has now. That might be a worthwhile feature request.

 

But I think you could get the equivalent result by rendering at 2x the size and resizing it down.

Posted
Multi-pass is just a primitive brute-force approach to AA and motion blur. The regular renderer is the one uses some intelligence to pick out things that need effort put into them. Perhaps it can be modified to offer the option of a 3rd anti-aliasing stage on top of the two it has now. That might be a worthwhile feature request.

 

But I think you could get the equivalent result by rendering at 2x the size and resizing it down.

 

I wouldn't call multi-pass brute force - just a different approach - Multi-pass will antialias all edges - including those created by decals (not sure about materials - didn't run test) - whereas the standard renderer excludes anti-aliasing of decal edges. The multi-pass method looks like an "averaging" of the passes of sorts. And the optional "soft" makes a big difference in multi-pass.

 

If ya don't want your decals antialiased - then use standard - much faster of course than 256 passes. But aliased decal edges will exhibit edge crawl in an animation for the "right" camera angles.

 

Eric's car doesn't look like it needs anti-aliasing of decals (not sure it has decals) - Perhaps go with regular renderer?

 

And of course rendering at 2x the size for a still works as well.

 

I used Robert's decal - purposely offset to create crappy edges - the resultant image was upscaled in photoshop - without resampling - just to more easily show differences in methods.

 

2nd imagine is just to show difference in using Soft versus NOT soft option with 5 pass multipass. Other images are just to show the difference in passes between 256.

big256vsaa.jpg

8_5softvs5NOsoft.jpg

2_256vs49soft.jpg

5_256vs9soft.jpg

9_5Softvs9soft.jpg

  • Hash Fellow
Posted

remember it's just the rectangular edge of a bitmap that doesn't seem to get AA'd in a regular render. The interior is AA'd just fine.

 

In actual practice the edge of a bitmap is rarely seen because most decals are created to be larger than the patches they are applied to.

 

If a bitmap has to be smaller than the patch it is applied to you can solve the edge problem by clipping 1 pixel in the alpha channel. And then you can use regular render and still have nice decals.

 

Here's five renders; the decal (with 1 pixel clip) looks equally good in all, but took way less time in regular render.

 

compositetestsalpha.jpg

Posted
remember it's just the rectangular edge of a bitmap that doesn't seem to get AA'd in a regular render. The interior is AA'd just fine.

 

I don't think that's true - if you look at first image I posted - the decal doesn't get AA'ed at all in the regular render, compared to the 256 pass render (and like I said before - I purposely made the decal ugly - used it straight from Robert's original jpeg post and skewed it)

 

http://www.hash.com/forums/index.php?act=a...st&id=44048

Posted
Here is a closeup of the rim. If you need another angle or an even closer shot, let me know. I hope it helps.

Thanks Eric. It does help because it confirms one my strongly held beliefs; if you start with the assumption that you're going to add as many splines as needed to fillet every corner you have nothing to fear from five pointers.

  • Hash Fellow
Posted
5 point patches are very stable these days which is just as well because they are pretty much unavoidable. ;)

 

Anyone remember the days before 5-pointers? There was something called the "hanging spline" that you tried to make work.

  • Hash Fellow
Posted
remember it's just the rectangular edge of a bitmap that doesn't seem to get AA'd in a regular render. The interior is AA'd just fine.

 

I don't think that's true - if you look at first image I posted - the decal doesn't get AA'ed at all in the regular render, compared to the 256 pass render (and like I said before - I purposely made the decal ugly - used it straight from Robert's original jpeg post and skewed it)

 

http://www.hash.com/forums/index.php?act=a...st&id=44048

 

I downloaded yours so I could duplicate the exact angle.

 

Here's what i get in a regular "final" render, multipass OFF. Blow it up and look at it. The decal is as anti-aliased as anti-aliased can be. It's even better anti-aliased than the the actual edge of the patch you can see beyond it.

 

FinalAA_NancyAngle0.jpg

 

here's a (hopefully) uncompressed png

 

FinalAA_NancyAngle0.png

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...