Welcome to the Second Life Forums Archive

These forums are CLOSED. Please visit the new forums HERE

Sculpt texture upload size

Gearsawe Stonecutter
Over there
Join date: 14 Sep 2005
Posts: 614
05-19-2007 11:12
If you are looking for as much accuracy as possible I'm finding your best bet is 128x128 right now to avoid problems cause by compression and the limit of only 256 values for each X,Y and Z. So small details on a large object sort of get lost or become chunky looking. A 256x256 map is sort of overkill. Because of compression I'm seeing a vert jump around +-2 sample points from where it should be at a 128x128 map size. Of course this is expected.

The image shows a 1 prim fence segment and 4 sculpt texture size at 32,64,128, and 256.
Haravikk Mistral
Registered User
Join date: 8 Oct 2005
Posts: 2,482
05-19-2007 12:57
Nice comparison!

Out of interest, how do they fair under level of detail? I'm presuming that if it down-samples from 128 x 128 to 32 x 32 that it's still passible as compression shouldn't cause any issues.
When you upload, are you using bitmaps (the default that the Maya script exports) or are you using jpegs? I was given to believe that once the images are converted to JPEG2000 they shouldn't lose much, if any, detail since Targa files are converted to it as well.
_____________________
Computer (Mac Pro):
2 x Quad Core 3.2ghz Xeon
10gb DDR2 800mhz FB-DIMMS
4 x 750gb, 32mb cache hard-drives (RAID-0/striped)
NVidia GeForce 8800GT (512mb)
Gearsawe Stonecutter
Over there
Join date: 14 Sep 2005
Posts: 614
05-19-2007 14:26
From: Haravikk Mistral

Out of interest, how do they fair under level of detail? I'm presuming that if it down-samples from 128 x 128 to 32 x 32 that it's still passible as compression shouldn't cause any issues.
When you upload, are you using bitmaps (the default that the Maya script exports) or are you using jpegs? I was given to believe that once the images are converted to JPEG2000 they shouldn't lose much, if any, detail since Targa files are converted to it as well.


I'm using uncompressed TGA files to upload from MAX. Again looking across the the tip of the fence post. you can see how one shade change in a color can effect the vertex placement. As the maps become smaller you can see how the color of one pixel affects another more. the 32 one is all over the place. The image below is wire frame mode in SL. It is not so much the size but how much each surrounding pixel effect another. Image compression is made for human eye. Take out the details that we cannot see or notice. there fore data is changed or lost. But now we can see it quite plainly when converted to geometry.



the total length of this fence is 4.25m at only 256 sample to work with that is about 0.016m of accuracy for each vertex at an object set to this scale. so the tip of this fence is about 0.05m wide. Yes I know this is tiny. And looks fine fore objects of this nature. but we can't limit ourselves to small texture maps. or a sclupty texture type with no compression. I really don;t know.
Gearsawe Stonecutter
Over there
Join date: 14 Sep 2005
Posts: 614
05-20-2007 09:29
One last test.
a 32x32 scaled to 128x128 using Nearest Neighbor in PhotoShop. This was to check my method in MAX was not screwing around with the color and proves how compression can alter the colors after it was uploaded.

http://www.joe-brown.net/SecondLife/SculptTut/sculpt_scale_32_to_128.JPG