Jump to content

Chodai Yoshikawa

  • Posts

  • Joined

  • Last visited

Posts posted by Chodai Yoshikawa

  1. Hi Wolf,

    1 Blender Unit is 1 metre by default.  there is also the option to work in metric or imperial and have the values displayed approprriately.    It's simplest to just leave it using Blender units I think as the unit display in Blender is not the best.

    In you Collada fils its easy to see what the scale the program that output the file was working in in the <asset> section of the file near the top.

          <author>Blender User</author>
          <authoring_tool>Blender 2.58.1 r39240M</authoring_tool>
        <unit name="meter" meter="1"/>

    read the meter="xxx" line. The value in that entry is the size of your CAD programs units in metres ie:   meter="0.01" means you units are in centimetres.  ie: 100 of your units ( cm ) to a metre.


  2. I use Blender so this may not apply in this case, but, perhaps you need to apply scale and rotations applied to your lower LOD and physics meshes.  ie: so all items on all LODS/PHYS have scale 1:1:1 and rot 0:0:0

  3. Hi Max, Asha,

    This is a weird problem and shows up in a few ways.   I have a file that uploads fine made with Blender 2.58 with the fix to change the bone naming that Tapple found.  This works fine with viewers in the 2.8 series using Joint Poistions  up to Second Life 2.8.2 (237914) Aug  4 2011 22:11:41 (Project Viewer - Mesh).   All the V3 series up to and including Fail using Joint Position in the upload with the same crazy offsets you see.

    The first image is the avatar uploaded using V2.8 237914 build.. and is superimposed on my avatar and the rigging works well.. The second image is wearing a version (from the same DAE file created with 2.58) uploaded using the current viewer.

    The second image is also what it looks like when created in Blender 2.59  r39475 and uploaded with EITHER V2.8 OR V3 viewer with OR without joint offsets and worn.  In all cases the rezzed only avatar in T-Pose looks fine.

    It seems to me, based on these observations,  that there is an issue with Blender 2.59 BUT also an issue with the V3 mesh uploader.

    AvViewer2.8 upload.jpg



  4. Hmm,

    With the current 3.0.1 viewer I cannot upload newly rigged mesh with Blender 2.5.9.  Nor can I upload previously uploadable rigged avatars without being distorted.  For a Mac based viewer at least using collada 1.4.1 files that *used* to upload fine its not working at all. :-(

    Uploading other mesh from Blender 2.5.9 using collada 1.4.1 files seems fine.

  5. I have found in the past that the uploader will upload only lower resolution textures compared to the original.  Perhaps its just picking the nearest lower size in standard powers of 2 or LL is trying to save on memory and data transmission costs. If you want the best results upload seperately and apply inworld, 10L an image is not much compared to the cost of mesh when it returns to its normal, higher price.    

  6. Hi Everyone!,

    A very good overall learning guide(s)  [PDF and also some video] for both the older 2.49b and 2.5 series is here:


    It is a guided entry into using blender and covers most things from simple control of the viewport up to video editing in a structured way.    It was meant to be used in the classroom, but can be used as guide to exploring blender by yourself.

    • Like 1
  7. Hmmm I originally had a single merged object as my physics mesh.  The 2 LOD levels I supplied both had 5 seperate meshes. I tried again with a physics mesh comprised of seperate mesh objects but had the same problems.  In either case both the Physics and two LOD meshes supplied had the same bounding box and centres.   I think I will try and reproduce with a simplier version of the meshes and file a JIRA.  It should work, and there is not documentation to say otherwise (although there *should* be some documentation to descibe the required relationship between physics and LOD meshes to make it work)

  8. Yes! I have the same problem.  After trying to load a custom physics shape the uploader removes the already loaded top and lower LOD levels.   IF I atttempt to re-load the top LOD shape it works BUT the physics shape I uploaded appears to be WITHIN a section of the re-uploaded mesh rather than the same bounds.   I could upload a similar physics shape before with the same mesh file but I added a few pieces to the Physics that went from a flat wall to an 3 sided room.  This has confused the uploader somehow.    I now have to upload the physics shape as a mesh and link it in manually inworld (making the physics shape 100% trans as well).  At first I thought I had a scale problem in my blender model but I made everything scale=1,1,1 in blender before trying again and no changes.   I have SLM turned off as well.

    Never got around to adding yet another Jira....perhaps I should.


    I experienced a similar thing on MSB#20.  I had just added some large linksets and moved a few things when I got the 5 miniutes to shutdown message. I logged and came back an hour later to find it was in the state I left if before adding and moving.   Perhaps the Sim only backs itself up on certain intervals and not just before a restart?

  9. Yes a much lower PE cost. ^_^

    I have two rooves, which were 26 PE and 6 PE (1k4 and 0k2 tris repsectively) both approx 28x18x5 metres with a custom physics shape.  When rezzed in MSB#21 dropped to 8 PE and 2 PE respectively.   This is looking much better for building now.

  10. Hi all,

    Step1 :  Lindens re-evaluate ARC cost taking into account Mesh costs, and true viewer load costs of sculpts and prims acurately.

    Step2: Lindens place a HARD limit on ARC ==> This promotes well crafted mesh as attachments with lower loads than current sculpt and prim use, which is inflationary and damaging viewer performance.

    Problem solved.

    Cho ^_^

  11. Hi all,

    Remember that the current head of the blender svn (>r38079) built with Collada enabled WILL export selected and works well for all NON RIGGED mesh. 

    Linden Labs have a fault with the mesh importer described in https://jira.secondlife.com/browse/CTS-660 that breaks rigged mesh unless you hack Blender to get around it.     If you want rigging to work please add a watch and vote to this issue to try and prompt the Lindens to look at it.    This issue is a general problem in the uploader with Collada and can possibly affect other tools outputs as well.

  12. Hi Charler,

    Yes, perhaps the wording as Cost rather than Tax is a better way of putting it.  

    In any case what can you say about the ARC? In particular ::

    * When is calculation being updated to account for mesh (I assume that sculpt and prim are correctly weighted now)?

    * Limiting the ARC level to promote sensible attachments (and encourage well made mesh)?

    Cho ^_^

  13. Hi Charler et al,

    I think most of us are aware that the prim/sculpty  == 1 Prim is not a true reflectionof server/sim/viewer cost.  The point is that because you do not want to break current users use of these you will not increase the cost of these items to reflect true load as you have done with mesh. This creates a bias of cost away from mesh and back to prim/sculpt where the true cost is higher than the accounting cost == Prim count.    It's like saying lets tax solar power but still subsidise coal.

    I hope that in upcoming ARC cost calculation changes you at least ensure that true costs are put into the calculation for all types : prim/sculpt/mesh to reflect an unbiased load.  Hopefully also you impose an ARC budget per avatar to promote the use of more cost (load) effective solutions to looking good.

    Cheers, Cho ^_^

    • Like 1
  14. Hmm yes me too... 151L$ for 20 odd tris with phys enabled.   1280Tris with auto LOD and using medium level as physics  PE=6 PH=3.32 ==> Cost to upload 159 $L....

    Uploaded 59k5 tris model, after having to do manual phys decompose, and using 7ktris for the lowest 2 LOD ==> PE =219 Phys 0.6 cost = 460 L$


    Seems 150 L is the base price.

  • Create New...