Noob question on projection mapping
polycounter lvl 16
Hey guys,
I really need some serious help.
I Have just finished modeling a very high detail car(1.8million tris) That includes the interior and exterior as well. The client I have done this for tells me that they want a 50,000 tri car built from this model.
Here is my problem, I have never done low-poly converting before This car has 785 objects in it. Man I don't even know where to start. Do I attach all of these objects to make it one and then just create a very low poly object around the car do some type of project from the high poly to the low poly version,If so how, Is there a tutorial somewhere that could teach me this.
Also,if this is the correct way, will the interior be projected as well as seen through the windows, or do I need to recreate the interior as well.
Please forgive my ignorance on this, but I am clueless when it comes to this type of thing and I am really strapped on time. They have given me 2 days to tell them if I can do this or not. So here's hoping you guys can get me going in the right direction.
Thank you so much for taking the time to try and help me.
-Brandon
I also wanted to add that the doors will not be opening the low poly model is basically being created to be seen from a distance, so there will be no closeups or even going inside to look at the interior as long as it can be seen from the outside
I really need some serious help.
I Have just finished modeling a very high detail car(1.8million tris) That includes the interior and exterior as well. The client I have done this for tells me that they want a 50,000 tri car built from this model.
Here is my problem, I have never done low-poly converting before This car has 785 objects in it. Man I don't even know where to start. Do I attach all of these objects to make it one and then just create a very low poly object around the car do some type of project from the high poly to the low poly version,If so how, Is there a tutorial somewhere that could teach me this.
Also,if this is the correct way, will the interior be projected as well as seen through the windows, or do I need to recreate the interior as well.
Please forgive my ignorance on this, but I am clueless when it comes to this type of thing and I am really strapped on time. They have given me 2 days to tell them if I can do this or not. So here's hoping you guys can get me going in the right direction.
Thank you so much for taking the time to try and help me.
-Brandon
I also wanted to add that the doors will not be opening the low poly model is basically being created to be seen from a distance, so there will be no closeups or even going inside to look at the interior as long as it can be seen from the outside
Replies
I assume you have 1 high poly model? If so it would be difficult to reduce the count without losing the shape, in this case it is always advisable to use meshsmooth (max terminology) modelling so you can see high levels of curvature that fit the objects shape perfectly while still retaining your low poly mesh.
As for texturing I would break the interior and exterior into 2 parts then recombine the maps into one. Along the way you will find you have to map parts independantly and then rearrange the uv's.
With any model I would say look at it's natural construction, break it apart then reconstruct it.
Yes, this in Max, and yes the high poly car is already created. I should have said, that both the high and low poly cars are going to be used, I just need to create a low poly car from this high poly one
As for the rest for your reply, Im clueless.
Any tutorials that you know that could help me learn this and am I going to have to unwrap the entire car?
accustudio tutorial: Decal Mapping of Low Polygon-Count Models
Low poly first & save file. Create high poly from low poly & save new file, project high poly onto low poly.
Yes you are going to have to unwrap the whole car, every face that is visible. It seems that a purely mental ray render and materials is out of the question so you will have to do this and substitute by using mental ray to create an ambient occlusion map for shadowing and then paint in or cut and paste the textures over the UV texture map to multiply as one layer against the ambient occlusion for that extra realism.
There are no tutorials that I can think of that would show you this process with your particular scenario as reducing complexity in a high poly mesh to make a low poly version is difficult when you want to create an accurate normal map.
The best advice I could give is to add a meshsmooth modifier then remove every other edge whilst checking with meshmooth on that the shape is still good. Keep reducing till you start seeing major changes in shape then you will know how far you can go to retain that detail.
Example : bonnet seams are high res details, be concerned with the outline shape and not the minor detailling.
What you want to do is create a lowpoly mesh over your highres mesh, you can load the highpoly into the background and create a new mesh over it, or simple take your highres and start optimizing it down to your target polycount. I generally do a bit of both. Then you want use max or xnormal to project a normals map and ambient occlusion map onto the low res, texture it, etc. Theres plenty of tutorials on how to create normal maps if you look for them
Thats a pretty backwards statement, you want to start HIGH and work down so you're not limiting yourself to the shapes of the orig lowpoly mesh, not the other way around man. You are limiting yourself MORE by "upscaling".
1. create high-poly version(done)
2. Clone each object and optimize each one for low poly
3. Unwrap all low poly objects
4. create normal maps using render to texture method
Does that sound about right? Also what about the windows im sure they have to be unwrapped, but im a little confused on if I need a normal map for them.
Last question, Would you recommend using polygon cruncher for this or is it better to just do it manually?
Thanks again for all your help
I really can't understand you saying that creating high poly first is the way to go as that would be a definite backwards process and not forwards progress.
I did mention later that reducing the complexity of the current mesh would be the best solution to actually ensure that the shape is kept rather than remodelling which would be less accurate.
Afterall the single high poly mesh available is the reason for the request of help, had it been low poly there would have been clear scope available and little confusion.
Each have their own methodologies - mine is to build the basic outline, cover all the contours and save the base mesh. After that add detail in stages, gradually refining the mesh till the end result is met for a high poly version.
You should try explaining this statement to the 80-90% of verteran artists on these boards that use this method then =P
You want the end result to be as close to your high resolution mesh as posible, so you do this first(well, after a simple block out) and match your lowres to the high. Its pretty simple really, what if you find that what you modeled in your lowres is lame and doesnt look good on the highres? If your lowres is already finalized you cant change your shapes. This is a very limiting process. When done creating the highres you should have the end result clearly in mind, its impossible to say you will have the end result clearly in mind before the highres is even created.
Yeah that looks good.
As far as #2, you should experiment a little. The two methods i generally use are these:
1. Use my sub-d cage as the base for a low poly part, and remove edges/optimize to get it to the final result
2. Recreate the shape from scratch.
I generally do both of these depending on the complexity of the shape etc. If i my sub-d cage for a certain object happens to be a mess, it can be faster sometimes to completely recreate than to optimize. So play around with this a bit and see what tends to be faster for you.
As far as using plugins to optimize, this can work, but generally will create a mess. So i wouldn't recommend it, but that doesn't mean you absolutely shouldn't try it.
Maybe I am lucky or just careful and avoid mistakes with low detail meshes, after a few years it kind of grows on you what works well is the way to go though I am open to new ideas and will give this a go sometime, when I have the time.
haha pior, neither on this one but still learning like everyone on Earth!
The intital model when I was done modeling was around 5 million tri's, after that I collapsed all my meshsmooths and started removing un-needed edges,which got it down to 1.8 million tris So basically what im going to do is go back through every object,clone them and start removing more edges again. Is that basically what you mean IronHawk
I'm not so sure I would waste time optimizing everything, it might be quicker and easier to just model out some new pieces around the existing high poly using tools like polyboost or Silo.
If you're under the gun and don't have the afternoon to get used to Silo, then polyboost would be a good option, its a great addition to 3dsmax in general. But really at some point check out silo, its so nice, so fluid, so fast, and so easy to use.
by adding 1 iteration the cage at least has some tension on the verts.
We all agree that the goal when doing a highpoly model is to be as fast and efficient as possible, since this is the heaviest part of the asset creation process. Now imagine that you want some sort of hardsurface organicish shape in your highpoly, like a funky futuristic gun handle. So you make a lowpoly*cage* for that, preview it smoothed, and push and pull verts to affect the smoothed version for it to look like the provided concept/or the idea you have in mind.
Problem is, when you look back at the lowpoly cage you will likely see spikes and even faces intersecting each other, because these act are control point. So if you use that lowpoly *cage* as a lowpoly *ingame mesh* it will not match the silhouette of the highpoly at all in some ares - even if it is it's source.
This is why most of the time, it's smarter to get a fast, good looking highpoly result and then figure out the bast way to built it's ingame counterpart.
Not saying that your method never works. It's just that if you stick with this method only you will run either into weird results (spikes) or a waste of time (trying to make the cage model work as an ingame model).
GC
http://wiki.polycount.net/Normal_Map#ModelingTheHighPolyMesh
I have been trying to get this rim ,normal mapped and it hasn't gone so well. I can't figure out what I'm doing wrong. I modeled the low poly version, It's encompassing the high poly one. I unwrapped the low poly model using the normal mapping. and then when I rendered to texture, its giving me a lot of black. Any ideas why? I'm not really sure what other information I need to give, so if there is anything else let me know. I uploaded two images onto my server of all my settings. I hope it helps and thanks for any info
www.billa1.com/images/pics/projection.jpg
www.billa1.com/images/pics/unwrap.jpg
-B
This is unheard of!
Also, I don't know what could cause that. I don't generally create my normalmaps in max. Have you tried xnormal, just to see if the problem doesn't lie with max? And you're absolutely sure the lowpoly cage encompasses the highpoly entirely?
Edit, i see your uvs now. Its hard to tell but are they in teh 0-1 range? If they're outside of that range they wont work correctly. Also your uv is set up a little strange, you should try planar mapping the front and back sides of that object and then just using relax on them to make sure the sides etc are getting a good amount of pixels. As it is now those lugnut holes you have would be stretching on your uvs.
Oh your referring to the the box in the uvw unwrap editor
http://area.autodesk.com/index.php/forums/viewthread/12885/#68318
www.billa1.com/images/pics/01space.jpg
It's going to be rendered in Max only, right?
So what are object-space maps?
http://wiki.polycount.net/Normal_Map#TangentSpaceVsObjectSpace
In Max's Render To Texture, go to the Projection Options dialog, and change it from Tangent to Local (IIRC).
Even if the single smoothing group gives it weird shading?
They haven't said anything about that, so I guess I should ask
Sometimes with tangent space maps you need to add extra geometry, or use smoothign groups because of the way they work, but with object/world space maps you dont have these limitations, they just *work*. There are some other technical limitations that you need to keep in mind when using object, so make sure thats what the client wants before you go all out and finish your model with those types of maps =P
I do want to say in the 5 years that I have been doing 3D, this is probably the most quick responded and courteous help I have gotten.
Thank you guys for everything so far and kudos to this site
here is an image of my uvs in the unwrap editor as well
www.billa1.com/images/pics/uvs02.jpg
Your uvs arent very effeciant as they are now, if you could upload an OBJ of them i could show you how i would lay it out real quick.
I just the reply email form the client who said they wanted all of them done with World. They also asked is it possible for to creat just a circle spline, collapse it to an editable poly and somehow or another, get an alpha map rendered for all the all the holes, is this possible? as well as a height map