Markus Altendorff <maa### [at] panoramasde> wrote:

> Well, i did test it, sort of, uhm... but not this time - the whole
> object group is copied straight from the IRTC "Escape" ("Escape^2")
> short i did (2003? 2004?). And it's not so much "flickering" as in
> "in-texture antialiasing too weak", but it's actually "jumping around"
> on the object... weirdest thing i've seen so far. It's OK for a few
> frames, then it shifts a bit to the left/right, then it's back to where
> it started. At first i thought i had accidentally animated some mapping
> parameter in a loop, but there's nothing in the timeline that would
> indicate such a thing, and i'm out of production time... i'll look into
> converting the cubes that build the doorframe to polygon meshes, maybe
> this'll help.
> The "usual" flicker of interpolation doesn't bother me as much because i
> compress the video to death anyway, so my problem is there's always
> worse MPEG artefacts than the grainy pulsing of texture edges :)

Not knowing anything about your application (and when has that ever stopped

had that a few times and when I re-create the file all has been well. If in
doubt blame someone else.

With my sort of animation lack of quality can kill it, there is no real

of space takes all the skill out of it.
Only kidding :-)

Stephen

Post a reply to this message

Stephen wrote:

> Not knowing anything about your application (and when has that ever stopped

> had that a few times and when I re-create the file all has been well. If in
> doubt blame someone else.

Heh :) but it was my mistake, of course... that, and the 
problem that precision is finite.

Mapping: cubical. My chosen angle: 45 degrees.
Problem: during rotation, the cumulated result down the 
object hierarchy may have been 44.999999 - or 45.000001
Result: every now and then, the texture was indecisive if if 
belonged to the front or the side of the cube.
Solution: use 44 degrees. And it's OK now. :)

I'll let it run again, so my video won't be ready until 
tomorrow afternoon...

-M

Post a reply to this message

Markus Altendorff <maa### [at] anthrosphinxde> wrote:
> Stephen wrote:
>
> > Not knowing anything about your application (and when has that ever stopped

> > had that a few times and when I re-create the file all has been well. If in
> > doubt blame someone else.
>
> Heh :) but it was my mistake, of course... that, and the
> problem that precision is finite.
>
> Mapping: cubical. My chosen angle: 45 degrees.
> Problem: during rotation, the cumulated result down the
> object hierarchy may have been 44.999999 - or 45.000001
> Result: every now and then, the texture was indecisive if if
> belonged to the front or the side of the cube.
> Solution: use 44 degrees. And it's OK now. :)


types of mapping :-)
Just out of interest how does cubical mapping work? I have never used it. I

mapping. I can see a line/edge on the cube giving problems but have
difficulty visualising a major problem. But what do I know :-)

> I'll let it run again, so my video won't be ready until
> tomorrow afternoon...


well.

Stephen

Post a reply to this message

Stephen wrote:


> types of mapping :-)

Hey, the doorframe is made from three cubes, so it seemed like a natural 
pick... waitaminute, it used to be "flat" mapping, maybe something was 
"corrected" when i moved it from C4D R8 to C4D R10... d'oh.

> Just out of interest how does cubical mapping work? I have never used it. I

> mapping. I can see a line/edge on the cube giving problems but have
> difficulty visualising a major problem. But what do I know :-)

It's no real "cubical" mapping as in e.g. a VR six-faced cube 
environment map, it's more of a "one image projected onto the faces of a 
"texture space" cube, and then mapped to the UV mesh of the real object."
I think it's because there's a tipping point where the "left side" 
becomes the "front side" if the surface normal crosses it, and with me 
defining the texture rotation exactly at that angle, any minor 
calculation rounding can make it flip to either side. Cool, my very 
first own fractal texture :)

> 
>> I'll let it run again, so my video won't be ready until
>> tomorrow afternoon...
> 

> well.

Unless my PC blows a fuse, it should work out. Need to tweak a few 
subtitles i'm not yet happy with, and just noticed that there's a few 
seconds where i mismatched the talk timing (A: moving mouth, B: 
gesturing...), need to re-run this. Another hour or so, at home. But 
since i don't need my workplace computer right now, i've launched the 
render there. Ouch. That box is SLOOOOOW. At home, the quad core eats 
through the frames at 30 sec/frm., this box here is more like 2 min/frm 
(single CPU G5 iMac).
I can definitely say, with my old home computer, i'd never have been 
able to do the film i managed this time :)

Oh, and in regard to encoding: I can hardly believe there's no decent 
and free MPEG-1 encoder for Mac... with the video editing stuff, i got 
all the shiny tools and the latest codecs, but MPEG-1 ...? "VCD standard 
150 kB/s" (waaay too big and wrong aspect ratio...) is all i can pick, 
with no manual overrides anywhere. So i'm transcoding it back and forth 
until i get an AVI-wrapped anamorphic DV-Stream, transfer that to my PC 
and run it through TMPEGenc Lite under Windows.

Well, at full resolution (720x360) it is a nice compact 29 MB H.264 MP4 
file, though. Maybe i should put that online somewhere, too ;)

-M

P.S.: Aw, so far there's nothing to be seen from Rusty? :(

Post a reply to this message

Markus Altendorff <maa### [at] panoramasde> wrote:
> Stephen wrote:
>

> > types of mapping :-)
>
> Hey, the doorframe is made from three cubes, so it seemed like a natural
> pick... waitaminute, it used to be "flat" mapping, maybe something was
> "corrected" when i moved it from C4D R8 to C4D R10... d'oh.



> > Just out of interest how does cubical mapping work? I have never used it. I

> > mapping. I can see a line/edge on the cube giving problems but have
> > difficulty visualising a major problem. But what do I know :-)
>
> It's no real "cubical" mapping as in e.g. a VR six-faced cube
> environment map, it's more of a "one image projected onto the faces of a
> "texture space" cube, and then mapped to the UV mesh of the real object."
> I think it's because there's a tipping point where the "left side"
> becomes the "front side" if the surface normal crosses it, and with me
> defining the texture rotation exactly at that angle, any minor
> calculation rounding can make it flip to either side. Cool, my very
> first own fractal texture :)





> >

> > well.
>
> Unless my PC blows a fuse, it should work out. Need to tweak a few
> subtitles i'm not yet happy with, and just noticed that there's a few
> seconds where i mismatched the talk timing (A: moving mouth, B:
> gesturing...), need to re-run this. Another hour or so, at home. But
> since i don't need my workplace computer right now, i've launched the
> render there. Ouch. That box is SLOOOOOW. At home, the quad core eats
> through the frames at 30 sec/frm., this box here is more like 2 min/frm
> (single CPU G5 iMac).

2 min/frame! How can you complain about that, luxury :-)

> I can definitely say, with my old home computer, i'd never have been
> able to do the film i managed this time :)
>
> Oh, and in regard to encoding: I can hardly believe there's no decent
> and free MPEG-1 encoder for Mac... with the video editing stuff, i got
> all the shiny tools and the latest codecs, but MPEG-1 ...? "VCD standard
> 150 kB/s" (waaay too big and wrong aspect ratio...) is all i can pick,
> with no manual overrides anywhere. So i'm transcoding it back and forth
> until i get an AVI-wrapped anamorphic DV-Stream, transfer that to my PC
> and run it through TMPEGenc Lite under Windows.




> Well, at full resolution (720x360) it is a nice compact 29 MB H.264 MP4
> file, though. Maybe i should put that online somewhere, too ;)
>


>
> P.S.: Aw, so far there's nothing to be seen from Rusty? :(

I thought that John said that he would enter something.

Stephen

Post a reply to this message

Stephen wrote:

> 
> 2 min/frame! How can you complain about that, luxury :-)

Tell me about it :) though what i've found has helped me most are the 
new graphics cards inside the new box, because i spend most of my time 
working "visually", and the abysmal speed of the old PCI/AGP mix i had 
before was really cramping my style ;)

Ah well, that see-through chainmail top really did kill the speed, it 
seems - with that out of the picture, even the older box is pushing to 
1/min.

> 



Well, inside the 3D software, it doesn't matter, and while editing the 
videos, it's much easier to handle all those files and folders than on 
the XP box i use for the occasional software development and the 
"compatibility"... - eh, looks? With a computer? It's below the desk, it 
could be a brick for all i care :) but sure is much more orderly on the 
inside of the chassis than my PC.

>> file, though. Maybe i should put that online somewhere, too ;)
>>


OK, i'll drop the URL once it's finished.

>> P.S.: Aw, so far there's nothing to be seen from Rusty? :(
> 
> I thought that John said that he would enter something.

Yes, that's what i remembered, too. But i keep on hoping... still 18 
hours to go, or something. It's UK-time based, isn't it?

-M

P.S.: Arrrgh! Need another 150 frames to add a final punchline... Well, 
not "need"... "WANT!"

*sigh*

Creativity, where were you half a month ago...?

Post a reply to this message

Stephen wrote:
> Markus Altendorff <maa### [at] panoramasde> wrote:
> 
>>Markus Altendorff wrote:
>>
>>>Hi, me again, sorry, ...
>>>
>>
>>and one more thing...
>>
>>how big can/should the thumbnail be? The current ones are square in
>>shape and rather compact - i've gotten used to the 320x240 of the old
>>IRTC...
>>
>>-M
> 
> 
> I had the problem the other way. How to make the preview intelligible but
> obscuring the nudity. I think that I made it 200x150.

> in the stage to metal which I like but the render time has gone through the
> roof 20 mins + time for a memory leak, per frame.

20 minutes per frame?  That's not long.  I generally don't do anything 
about render times until they're over a half hour.  The memory leak is a 
problem, though.

Regards,
John

Post a reply to this message