Octane discussion (branched) Closed  1-20  21-36
Thread Split: Some posts in this thread were moved from here

Next
 From:  OSTexo
4415.1 
Hello,

Octane still has serious problems with normals, and the developers refuse to acknowledge it. TheaRender is very good, it's not a GPU based solution but gives great (as well as accurate) results and the caliber of development seems to be much higher.

EDITED: 11 Aug 2011 by MICHAEL GIBSON

  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Mike K4ICY (MAJIKMIKE)
4415.2 In reply to 4415.1 
Thanks for the insight OSTexo,

Yes, I could tell that Lux employed some of the same techniques as Thea, such as MLT and BiPT.
I'm not sure if there are any major fundamental differences between Kerkythea and Thea Render, besides the engine.

GPU use would be nice, I have a GeForce 6600 and the slots on the board to set up and SLI. At least MoI runs decently.
(..that is, once I asked for "tripple buffering" for IE in the NVidia console.)
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  YHWH_777
4415.3 In reply to 4415.1 
<< Octane still has serious problems with normals, and the developers refuse to acknowledge it.>>

I read the Octane forums almost every day, and I never see any complaints like the one that our accusing them of. About a year ago, there were some issues with normals when the software just came out, but those were promptly fixed. I regularly model in MoI and Blender and render in Octane and never have problems.

The only problem that I have right now is when I try to export/import from MoI to Blender. The latest versions of Blender (2.5x) have problems with MoI normals. Hopefully someone will write a fix.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  BurrMan
4415.4 In reply to 4415.3 
"""""""""About a year ago, there were some issues with normals when the software just came out, but those were promptly fixed.""""""""

Pretty sure that the guy writing the software came here and outright denied there was any issue and told the user to shove off...

Maybe you can revisit that thread and do the "test" to see if they are still throwing away/overwriting/ignoring the original vertex nomals for the NURBS data...
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  YHWH_777
4415.5 In reply to 4415.4 
<< Pretty sure that the guy writing the software came here and outright denied there was any issue and told the user to shove off...

Maybe you can revisit that thread and do the "test" to see if they are still throwing away/overwriting/ignoring the original vertex nomals for the NURBS data...>>

Yes, there was an issue quite a while back. But it was fixed. Like I stated, I use MoI and Octane on a regular basis. There are no normal issues anymore.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  BurrMan
4415.6 In reply to 4415.5 
YHWH_777,
I was hoping you could revisit that thread and do the simple test presented, and return the result for this forum. The developer of the software had stated that there was no issue, never an issue and nothing would be changed...
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4415.7 In reply to 4415.6 
Hi Burr, I'm pretty sure that some point after he said all that it did actually get fixed though.

I haven't heard any problems mentioned about it for quite a while after that so I think that it's been fixed for a while now.

It definitely did take quite a lot of repeated explanations and various tests sent and stuff like that until it finally got fixed. I seem to remember that some tests with the sample objects verified that it was fixed at the end of all that.

- Michael
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  OSTexo
4415.8 
Hello,

Maybe the thread was deleted, I'm not sure. I do know for a fact that there are continued problems with normals and Octane. I did various tests on both very simple and more complex objects, loading them into Thea and Octane and Octane always had the problem, and the problem was consistent, it just was more obvious in some models than others.

I'd like to like Octane, but they really need to fix the fundamentals before worrying about little feature improvements here and there. Until they get this fixed I can't see Octane being used by anyone who cares about their work, although it does do a decent job as a quick visualization tool, although not accurate. Here are the two sets of images, Something is seriously wrong with Octane, Thea looks good, don't you think? When I presented these images for some sort of explanation to the developers I was met with silence. I also provided the OBJ files, I got all sorts of doubletalk from Octane users about what could be wrong, anything from double sided polys to not subdividing enough, but I never got an explanation that made sense or that wasn't contradictory. I'm still waiting for a response from a developer, but I'm not holding my breath.

I have to say I am much happier with Thea, the developer took the time to respond personally to one of my requests for information, which just solidified in my mind that when you deal with Solid Iris (creators of Thea), you are dealing with a standup company.
Image Attachments:
Size: 88.1 KB, Downloaded: 117 times, Dimensions: 1834x600px
Size: 166.6 KB, Downloaded: 124 times, Dimensions: 2058x768px
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4415.9 In reply to 4415.6 
Hi Burr, here were the posts with those test files verifying the fix:

http://moi3d.com/forum/index.php?webtag=MOI&msg=3472.82
http://moi3d.com/forum/index.php?webtag=MOI&msg=3472.87

Here's the one mentioning the fix:
http://moi3d.com/forum/index.php?webtag=MOI&msg=3472.79


That was more than a year ago, so unless there's been some regression that problem has been fixed for a while now.

- Michael
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  OSTexo
4415.10 
Hello,

By the way, those renders were done with the latest stable release of Octane as well as Thea.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  BurrMan
4415.11 In reply to 4415.9 
""""Hi Burr, here were the posts with those test files verifying the fix:""""""

Thanks Michael,
I fell asleep and missed those.... It was left hanging in my mind.. Thats good news.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4415.12 In reply to 4415.8 
Hi OSTexo - maybe it's something to do specifically with translucent/refractive materials, which seem to be used in both your examples there?

Also when exporting from MoI try using "Output: Quads & Triangles" instead of "Output: N-gons", just to avoid any potential complication from processing any complex n-gons by the renderer.

- Michael
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4415.13 In reply to 4415.8 
Hi OSTexo - the other thing I seem to remember is that it didn't handle the situation of 2 faces sharing a point location in common but having unique vertex normals for each one. You'll get a mesh like that if you have the "Weld vertices along edges" option set when you are exporting from MoI.

So I'd recommend turning welding off and set "Output: Quads & Triangles" when exporting to Octane and see if that makes any difference.

- Michael
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  OSTexo
4415.14 
Hello,

Both were exported triangles. I tried the other options including n-gons but ran into the same problem. You are correct that if the object is not transparent / translucent it doesn't seem to exhibit these problems. From a purely consumer standpoint, this seems to be a pretty fundamental issue for a renderer, how you can ignore the elephant in the room makes me wonder about the caliber of development going into Octane.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4415.15 In reply to 4415.14 
Hi OSTexo,

> Both were exported triangles.

That should be fine - either all triangles or quads & triangles will avoid any n-gon processing issues in the renderer.

But what about welding? Did you export from Moi with welding on or welding off?

At least at one point in the past Octane had difficulty dealing with welded meshes, if you exported with welding enabled then that could be what you were running into.

- Michael
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  OSTexo
4415.16 
Hello,

Exported from MoI ,quads and triangles, welding off, the Octane Render doesn't look to good. This will render correctly in Thea.
Image Attachments:
Size: 40.8 KB, Downloaded: 83 times, Dimensions: 1024x512px
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  YHWH_777
4415.17 In reply to 4415.16 
OSTexo:

I believe that those are smoothing issues and not normal issues. If you make your mesh finer (by increasing the number of triangles), then it should work much better.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  OSTexo
4415.18 
Hello,

So how do you explain the smooth result in Thea with the same file? The vase objects are certainly dense enough to show a proper result. It's a bug in Octane, a finer mesh does not fix the problem.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  YHWH_777
4415.19 In reply to 4415.18 
<< So how do you explain the smooth result in Thea with the same file?>>

Thea and Octane are extremely different programs and use different smoothing algorithms. IMO, one of the weaker areas in Octane is that it requires a very high polygon count when using glossy or specular materials. Thea actually announced that they had rewritten their smoothing algorithms to get rid of the terminator artifact issue that many other renderers have. You can read about it here: http://www.thearender.com/forum/viewtopic.php?f=17&t=4331

<< The vase objects are certainly dense enough to show a proper result. It's a bug in Octane, a finer mesh does not fix the problem.>>

I seriously doubt that it is a bug in Octane. There are hundreds of amazing images that are coming out of Octane every month. Nobody on their forums is complaining about this issue. Why don't you post your issue on their forums. There are lots of friendly users that I am sure would be glad to help.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  OSTexo
4415.20 
Hello,

The terminator artifact problem is present in Octane, I tried some tests thinking that might be the problem, but it doesn't make sense that when I switch to different materials in Octane the problem goes away using the same model, does it?

I'll have to disagree with you about it not being a bug in Octane. Like I said previously I have posted the issue on the forums, and have heard nothing from the developers. I get the indication that the developers are aware of the problem, but choose to ignore it. In fact, there are numerous examples on the Octane Render forums with similar faceting in transparent materials present in objects created in modelers other than MoI. Just because someone doesn't complain about a problem doesn't mean there isn't a problem. Some of the more vocal users on the forum gave the same lines that you are typing, not enough mesh density, smoothing problem, problem with the model, which leads me to believe these users are not interested in insisting on a fix, they choose to ignore the error as well.

In short, this problem is not user error, it is an Octane error. I realize it is in beta, so hopefully someone over there will come to their senses and squash the bug causing this. I'll agree with you that it is a fast renderer, although not accurate at this point. I think the Octane developers are working at a significant disadvantage. I have to think that the Thea developers have a much greater depth of experience in this field, so it is not surprising that they are much further ahead with a relatively slower CPU based, but highly accurate product.
  More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged
 
 

 

 
Show messages:  1-20  21-36