View Full Version : VW Beetle retopology
12-30-2011, 01:25 PM
Hey guys. I'm starting to make a game-ready version of a Beetle model that I did a while ago that was originally intended as a cinematic prop. I haven't done much of this before and I was hoping for some pointers on workflow.
At the moment the model consists of about 30 separate subdiv objects. I'm thinking I'll delete support loops, optimise, then weld these pieces together to form a main 'chassis' object around which the remaining elements can float.
Then I'll re-do the low poly UV's and bake normals, AO, diffuse and specular to the low poly (to avoid manual repositioning of texture parts in photoshop. downside to this is I'll lose the ability to edit my file in Photoshop)
Something that's confusing me: I'll be baking the high poly geometry to the low poly, but the high poly also has detail normal maps applied to it; how can I preserve this detail in the new normal map?
So does this sound like the right way to do things? What would be the approximate tri count for a prop like this?
Here's the original model:
12-30-2011, 01:43 PM
2-5k tris would do.
12-30-2011, 02:11 PM
I can't tell how much detail you got inside the car, but in any way, if you want to keep a certain quality, 2000 to 5000 triangles sounds very optimistic.
I usually go for "as much as needed, as little as possible".
As far as I know, xNormal allows to add normal maps to highpoly meshes, I've never tried it myself, but it sounds like what you need..
12-30-2011, 02:24 PM
If it isn't actually going into a game then don't worry so much about triangle count. 5k is fine but if you go a little higher it will stay relevant in your portfolio for longer.
12-30-2011, 05:29 PM
I wouldn't sweat the numbers as much as I'd worry about preserving silhouette.
01-08-2012, 01:15 PM
Thanks for the advice guys. I should add that the reason for doing this was so I had something to submit to Blitz studios in my application for their open day.
The model ended up at over 10k tris after heavy optimisation, mainly due to the way I'd done the original cinematic model with interior surfaces for every piece of bodywork. I wanted to fix this by connecting up the geometry, re-unwrapping and re-texturing the whole thing but unfortunately didn't have time. Probably could have halved the poly count that way!
Anyway I learnt a lot about how to optimise and how NOT to bake normal maps during this process. The result looks okay, but the normals are messy; next time will be better :)
Here are the sheets I submitted to Blitz, any crits on the model/textures or presentation are welcome.
Why are there so many tris on the wheel mounts? Those tris could be doing a lot more for the cars body, giving it a smoother silhouette.
01-08-2012, 01:29 PM
First thing I thought of when I saw it was what ZacD said as well.
01-08-2012, 04:12 PM
Wow, can't believe I didn't notice that discrepancy, it always helps to get some extra eyes on my work. Gotta be careful when modelling elements in isolation!
01-09-2012, 12:59 AM
Despite the body/wheel poly density discrepancy it does look pretty nice. If you had time to clean up the bakes and transfer those poly's to the silhouette it would a great piece.
How did the open day go?
01-09-2012, 06:27 AM
Yeah those wheel hubs are eating a crapton of polies. Looks like they're close to 1000 tris apiece, so the four of them are about a third of the polycount of the whole car.
Especially wasteful since in racing games they are one of the few elements of a car that could benefit from normal maps (instancing + small size + harsh detail instead of smooth flows) instead of using modeled curves. I know this isn't exactly in racing conditionm but still...
01-10-2012, 11:56 AM
Thanks ysalex. The open day is not until March, but Blitz needed a work sample sent to judge my worthiness :) I'll find out if I got a place on the 23rd.
Snader, I did some further optimisation on the wheel hubs and got them each down to 266 tris from 571 originally. Still too high, and I could probably halve that again by baking some of that detail into normal maps as you suggest, but I'll have to leave it for now.
vBulletin® v3.8.4, Copyright ©2000-2013, Jelsoft Enterprises Ltd.