PDA

View Full Version : What's a good workflow to model different LOD meshes?


Oli4D
02-21-2007, 04:39 AM
Hello everybody!
I'm not a XSI User yet, but I'm thinking of learning and using it. I heard a lot of good things about it, especially about the modeling and animation tools.
I'm unfortunately quite far away from home (on the other side of the world ;-)) and can't test anything... or try the demo.
So that's why I ask you guys for help.
It could be that I soon get a new job where I will have to do a lot of modeling and texturing for real time stuff.
It will be necessary to model and textur airplanes and cars in four differnt LOD: 300, 600, 1200 and 2500 polygons.
My question is: What would be the best workflow to do this in XSI? (I might use Bodypaint3D for texturing).
Especially what would be the best workflow with XSI 6 Foundation... because I probably will not be able to afford 'Essential' (with the poly reduction tool).
What would you guys recommend? What is your workflow for tasks like this?
I thought about modeling and texturing each of the four versions individually... But maybe there is a better solution.
What would you recommend? If you need more information to answer this, please let me know as well.
Thanks very much!
Oliver

Mic_Ma
02-21-2007, 07:14 AM
In general: Make the high poly versions first, then remove/merge/weld points/edges/polys in stages for the different LODs. Keep your UVs sensible and non-overlapping, if possible. Same goes for clusters, geometry, naming, etc. Check if different textures are used for different models. Check if the game engine can tile textures.

Triangulate at the end, where necessary.

Chris-TC
02-21-2007, 03:42 PM
In general: Make the high poly versions first, then remove/merge/weld points/edges/polys in stages for the different LODs.

I don't agree, but that's probably down to personal preference.

I model almost anything starting with a box. Then I add edge loops and gradually increase detail.
Therefore, I'd start out with a box, put in as much detail as I can with the 300 poly limit in mind. Then move on to 600, 1200 and 2500. My normal modeling workflow wouldn't even be disrupted this way.

mdee
02-21-2007, 05:06 PM
I do it like Chris. Model low res then go up with details.

Oli4D
02-21-2007, 09:50 PM
I don't agree, but that's probably down to personal preference.

I model almost anything starting with a box. Then I add edge loops and gradually increase detail.
Therefore, I'd start out with a box, put in as much detail as I can with the 300 poly limit in mind. Then move on to 600, 1200 and 2500. My normal modeling workflow wouldn't even be disrupted this way.

Sounds logical.
My objects will be Airplanes and Cars.
How do you deal with the texturing? Apply separat UV for each mesh Version, or what would be the best workflow for that?
I'm thinking about using Bodypaint3D for that because I already know it quite well... what do you think about that?

Anyone else with Modeling Workflow Ideas for diferent Versions of the same Object?
Thanks very much, Oliver

Chris-TC
02-21-2007, 10:38 PM
I guess the textures should be pretty much identical for all versions? That's an ideal situation for GATOR, but unfortunately it's not part of Foundation.

In this case I really can't think of a good way to do it other than unwrapping all versions somewhat similarly by hand.

Mic_Ma
02-22-2007, 12:05 AM
Low → High if you are confident of the designs.

In practice I find it easier to do the High-res first since it is the most important model and any changes can simply trickle down. The very lowest res is often so simple it can be done in minutes, from scratch. Otherwise you might have to do the whole series again and again, everytime the art-director/boss/client/somebodyelse changes their mind. I rarely start from a box. More often than not I get/find some "leftovers" from the past, other-team-member/company.

But yes, if it is straighforward workflow. Low → High sounds good.

Oli4D
02-22-2007, 02:30 AM
Low → High if you are confident of the designs.

In practice I find it easier to do the High-res first since it is the most important model and any changes can simply trickle down. The very lowest res is often so simple it can be done in minutes, from scratch. Otherwise you might have to do the whole series again and again, everytime the art-director/boss/client/somebodyelse changes their mind. I rarely start from a box. More often than not I get/find some "leftovers" from the past, other-team-member/company.

But yes, if it is straighforward workflow. Low → High sounds good.

Yeah I can see you point, thanks very much for writing that.
But in my case it will be the modeling of existing objects such as real Airplanes (for ex. 747) or existing cars. So, there probably won't be any design changes.

Oli4D
02-22-2007, 02:38 AM
I guess the textures should be pretty much identical for all versions?

Yes, as usual :-)

That's an ideal situation for GATOR, but unfortunately it's not part of Foundation.

In this case I really can't think of a good way to do it other than unwrapping all versions somewhat similarly by hand.

Hmm yeah. The problem is: I will have to buy the software for my own, and I simply can't afford the 'Essential' at the moment. I don't really need it... or well... maybe I do...
But yeah know, I'm really looking for a solution with the foundation, especially because nobody ever mentioned the 'poly-reduction' of the Essential. It seems not to usual to use that feature. One reason more to choose 'only' the 'Foundation'.

Are there any other tips oder hints? Is anyone of you guys using Bodypaint3D for texturing?

The unwrapping all versions similar by hand would work I guess...

JurajMolcak
02-22-2007, 06:00 AM
I guess the textures should be pretty much identical for all versions? That's an ideal situation for GATOR, but unfortunately it's not part of Foundation.

In this case I really can't think of a good way to do it other than unwrapping all versions somewhat similarly by hand.

Hm, I didnīt try that, but maybe it could be good to model whole bunch from low--->high, than texture and normal/bump map the highest one and than do "ultimaper stuff" for the rest out from that highres model. You donīt even need to have UVW setting for lower versions, as far as I know, coz ultimaper will generate that. So (theoreticaly)(might try that later this week):
1. Start out from BOX with polycounting in mind
2. Send HIGHEST version to bodypaint
3. Make UVs and textures/bumps there
4. Send back to XSI FND
5. Make ultimaping from HIGH 2 LOWER and lower and l o w e r (workflow is easy, nicely discribed in manual)

Oli4D
02-22-2007, 09:48 PM
Hm, I didnīt try that, but maybe it could be good to model whole bunch from low--->high, than texture and normal/bump map the highest one and than do "ultimaper stuff" for the rest out from that highres model. You donīt even need to have UVW setting for lower versions, as far as I know, coz ultimaper will generate that. So (theoreticaly)(might try that later this week):

1. Start out from BOX with polycounting in mind
2. Send HIGHEST version to bodypaint
3. Make UVs and textures/bumps there
4. Send back to XSI FND
5. Make ultimaping from HIGH 2 LOWER and lower and l o w e r (workflow is easy, nicely discribed in manual)

Thanks for your posting Oweron!
Yeah that sounds interesting. Ultimaper is part of the Foundation?
It would be great if you could test it later this week. Would be interesting.
I can't... because I'm in Australia at the moment.

RmachucaA
03-12-2007, 05:41 AM
It comes down to personal preference.

But i too start with the highest detail, then work down through the lod's. Why you may ask?

Simple, i want to keep the lower lods as CLOSE and DETAILED as possible as the high lod.

Oli4D
03-15-2007, 08:16 AM
It comes down to personal preference.

But i too start with the highest detail, then work down through the lod's. Why you may ask?

Simple, i want to keep the lower lods as CLOSE and DETAILED as possible as the high lod.

Heyho!
Sounds interesting, thanks for the post!
Well and how are you actually working your way down? Modeling each modeling again completely new, or use the hi-res mesh and delete / merge polys?

el_diablo
03-15-2007, 09:14 AM
Also clusters et al don't propagate correctly through adding detail. They act predictably during lowresing tho. Keeping the shape close to high res would be another reason too.

bryann
03-15-2007, 09:57 PM
I much prefer going high to low, as well. It's much easier to keep the silhouette right and also easier to see which details can go based on how well the texture is selling that lost geometry.

CGTalk Moderation
03-15-2007, 09:57 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.