Michael > I think client side smoothing would help with your issue too, as it offers endless LOD possibilities. E.g. use a low.poly wheel at distance (8 sided), the 3 iteration smoothed one close up.(64 sided) You could also adjust the LOD depending on device, using the same assets, simply by monitoring the actual FPS and/or screen size. There's a paper by the PS2 team on this. (will try and find the link...)
Richard > I'm going to sound like a 40 something who still remembers modems (which I am), but the only time I ever think of bandwidth as "not a problem" is working locally :) Mobile internet means this issue will be with us for a long time to come, (bandwidth and charges) not to mention device memory capacity. If I can shave 10% off filesize, without deterioration, I would generally do it. Shaving 90% off has got to be interesting hasn't it? If another file format is smaller to start off with, then even better. (had a few probs with .3ds in Broomstick and found .OBJ more stable i.e. models loaded every time without normals flipped/facets etc. will test again with latest build) If it should prove to take 10 minutes to run a smoothing algorithm in Flash, then I'm totally with you, but I can't imagine that will be the case. In a java example I've seen it's almost instant. I'll start with an AS3 implementation, just to get familiar with the maths, but if that's too slow, maybe a PixelBender shader would be the answer? For my current project scenario it would be a one time operation when a model first loads, with fallback to low poly when viewed at distance, or with many objects in the scene. I will gladly share my test results here, and would be delighted to contribute to this fabulous engine :) Skål! Pete On May 23, 6:53 pm, richardolsson <[email protected]> wrote: > It's an interesting idea, but bandwidth is not really as big a problem > as you would think if you choose formats that are more optimized than > OBJ. Looking at geometry only, 3DS tends to be about 40% of an OBJ > containing the same data, and deflated AWD2 is about 20%. So if you > would use AWD (as will be our recommendation once Away3D 4.0 and AWD > is stable), you would likely be able to store the smoothed model in > 250kb, which is not that bad. > > Still, it's an interesting idea, and I would be keen to see some > benchmarks on how quickly the smoothing can be calculated in AS3. > Simple sub-division is of course rather quick, but smoothing is likely > a bit more complex and I'm not convinced the actual operation of > smoothing the geometries would be fast enough (especially on lower end > CPUs) to provide any substantial improvement over loading high-res > models in optimized formats. > > If you are interested in trying it out and contribute the code, let us > know, and we can discuss it! > > Cheers > /R > > On May 23, 1:40 pm, Michael Iv <[email protected]> wrote: > > > > > > > > > Adding more subdivisions is not the biggest problem here as it seems to me, > > but rather that Molehill is not so strong as native OpenGL to > > run smoothly such high res models .Anyhow would like to see such a feature > > in Away3D :) > > > On Mon, May 23, 2011 at 2:04 PM, themightyatom > > <[email protected]>wrote: > > > > Now we've finally got the power to display high-res models in Flash > > > via Molehill, the next bottleneck seems to be bandwidth. > > > In 3D Studio MAX I am often using TurboSmooth on relatively low-poly > > > models, taking them from 100Kb, to 1.2Mb when exported as .OBJ > > > > It struck me that if we could implement TurboSmooth, or similar > > > smoothing algorithm in the client, we could send low res models over > > > the wire, and have them come out smooth in the browser. > > > > Is anybody working on this in Broomstick? I've got loads of examples > > > of the "theory", but the only source I can find is OpenGL/C++ (and > > > pages of it) > > >http://www.holmes3d.net/graphics/subdivision/ > > > > I will be looking into Loop Scheme myself, but any pointers, examples, > > > help, etc. greatly appreciated :O) > > > > Cheers, > > > > Pete > > > -- > > Michael Ivanov ,Programmer > > Neurotech Solutions Ltd. > > Flex|Air |3D|Unity|www.neurotechresearch.comhttp://blog.alladvanced.net > > Tel:054-4962254 > > [email protected] > > [email protected]
