Soft modeling (organic)
-
@kwistenbiebel said:
I am still convinced that trees a la Xfrog, but 10 times less polygons than that collection, would make a great commercial package.
+1
My PC thinks it's a 286 when I load even just 3 Xfrog trees. -
Thanks guys, I agree there is enough of this kinda stuff available.
I will have another look at the tree thing, 5mb sounds pretty doable, but I will need to try a few ideas, I may need to use .png's for the foliage however, which gets complicated with apps like Vray that make using png's so awkward.
If I was to create these trees would making a basic low polygon shape named say 'Birch-proxy' accompany the high poly 'Birch' so that one can do the instancing in SU? -
Solo, check out these trees ande. made. (http://asgvis.com/index.php?option=com_smf&Itemid=90&topic=5508.15) They honestly are the best trees I have delt with. They arent as cumbersome as xfrog (which end up just crashing my comp. anyways), and they look just as good in my opinion. But they use alpha masking, and they work fairly well. I think the trick with v-ray is to use the same file type for the actuall texture and for the transparency mask, because I have tried using a jpeg for the texture and png for alpha, and got bad results. So I usually take the png. file and make my texture and alpha in PS and then apply them to the component, foilage. I have fairly strong computer I was using but was able to get about 15 trees in my scene and I could have probably fit a fair amount more. They really were the best solution I have seen for SU. So I would try to encourage this technique with the model and png. texture included, but it ultimately is up to the artist. Either way I know you will give us something great. Keep an eye on the gallery, I am going to try to get the renderings up that I did with these trees sometime soon to give you a better idea of how they turn out.
-
@solo said:
I will have another look at the tree thing, 5mb sounds pretty doable, but I will need to try a few ideas, I may need to use .png's for the foliage however, which gets complicated with apps like Vray that make using png's so awkward.
Using PNG for foliage is ok. It's not a big deal to add an alpha channel. But it's nice of it isn't built up of too many different materials as currently V-Ray material won't import with components so the process of adding the alpha channel have to be done each time.
An XFrog trees often comes in with between 80K-120K faces, which quickly kills a model. But it'd be very interesting with <10-20K trees that still look good.
XFrog usually have each leaf, or maybe 3-4 leaves in a png, but it's still allot. A lib with trees that's lighter, but still appear detailed, with volume would be very interesting. -
@solo said:
Good idea, I will add about 3 variations.
Hi Solo how do you map these meshes like that... its not done in SU i think... im i correct.. id like to learn how to map organic shapes
-
rombout, It's 100% Sketchup, created and textured in SU.
I do not have a tutorial on hand but keep an eye on the main site where the new tutorials are, I am sure there will be a few organic modeling tutorials coming soon.
-
I think this will be an excellent package - not everyone is aware how good these soft furnishings can look,
would certainly add realism to interior shots.
I'm always amazed by your modelling -
-
Solo, you are a great talent and I have learned much from you. Thanks.
The problem I have with most packs is there size.@unknownuser said:
so far that are medium poly at about 600-800kb.
I don't consider that to be a medium mesh in SU. I really am tired of hiding these objects in another layer until it's time to shoot....using proxy's, splitting symmetricals, etc...etc. Too many hoops for me. I like to be able to see as much in context as I can while modeling without spending hours preparing them.
If you could produce a set of useful objects that are light, I believe they would sell.
-
What do you consider a medium mesh?
-
My opinion is subjective I will admit based on my modeling behaviour (fever) and content. I also need to be able to convert to solid (in 70% of the cases) which is proving to be possible only with careful model construction.
Regardless, a medium mesh in my libraries would be about 200K. -
-
@unknownuser said:
200kb medium? wow! that's low in my book, so what do you consider low and do you have any samples of medium and low components that look half decent?
Less than 100k. I'm not going to try and qualify my advise. You know yourself this is a perpetual problem and balancing act between pre-vis quality and model size. If you have the time to put together packs and you are interested in earning a living then I will leave it to you to create decent low poly models. You have got the visualization down, you are a master. I would turn my attention to building effective low poly packs. If this is something you can achieve, I do believe that this approach would help you gain some momentum.
If you are running out of ideas then I suggest you try to build some solids with SU. What you will see very quickly is the basic primitive SU tool set is still most efficient. You will be shocked by the messy, hole laden models that the SU power tools produce, i.e. follow-me, push-pull. This is even more pronounced when you rev up these tools with Rubies. What you are faced with is hard work to achieve what I suggest. You would need to set asside the power tools and see exactly what can be built to your standards with the least number of faces and edges in SU.
-
Solo....I hope this gives you some ideas.
There are many complaints about the speed of SU and its inability to handle large poly counts. In my engine development world which has to remain nameless, the ability to carry heavy models with C++ code is based on the accuracy of the models in the world space. SU is a surface modeler designed for ease of use. It has features that require very complex algorithms, inferencing a good example.
These algorithms are built to make assumptions about the real world space. The placement of faces and edges in Sketchup cannot be described as accurate. As a result, the algorithms need to be very complex and take up a lot of processing time to make assumptions about how the mesh is connected and where each wire is positioned in the real world. As it is, SU does a better job of this than anyone. That is why it is so easy to use.
If you have access to engines that can recognize solid as well as surface models (for the purpose of accurate measurement)in the same environment you will find that the solid models are very easy to work with. Their geometry is true. The algorithms don’t need to churn through hundreds of tests on the geo in order to understand its present state and position. The state is a constant, it is accurate.
On the other hand, if you run an identical model that is not solid, there is a large computing overhead required just to figure out the true shape of the model let along its true position in the world space.
Before Sketchup can move ahead to heavy models and complex meshes it needs to reduce the complexity of the algorithms required to support complex surface models by becoming more accurate in laying down the mesh in the first place.
This is a major undertaking as we are finding out in my little part of the modeling world.
What you can’t see in Sketchup is how inaccurate the mesh really is. It looks perfectly normal. I don’t see any holes in your couch above Solo. On close inspection you will find that it is a mess. It has nothing to do with your modeling. It is a product of the tools. An easier way to explain would be to say you are not seeing all the hidden geometry. You are not seeing the geometry that the C++ algorithms see.
The (proprietary)engine that I use reveals a lot of secrets about real mesh. For example, I can take two models, both modeled in Sketchup, both look the same and the file size is the same. The mesh looks exactly the same (in Sketchup) as well. One of the models was built using the line tool only. The other with all available native tools in Sketchup. Even though the models are materially the same the model built with the line tool can be read, understood and moved with less than half the computing overhead. It has more accurate and consistent real world coordinates and so it is much easier for the C++ to digest without wild algorithms bordering on AI.
What you see is not what you are getting. In order to build more efficient models in SU you need to be able to see the mesh as the C++ would. -
I got about 1/2 of that . ..
-
Well I am sorry it was a bit painful. I was a few years wondering about this stuff without a clue until I was forced into the solid modeling world. For me/us it is about "what you see is not what you get". Although surface models look great it is what you cannot see that is the problem. It is not an easy subject for me but it has become necessary to understand the differences. As you can understand it is not possible to fix something if you cannot see it. For me the whole exercise has been a revelation in my model building career.
-
NOt at all. Im just a dummy when it comes to this stuff.
I do model quite a bit of Drapery as per Solo's technique and then when I have to make it different I use Freescale and/or FFD to reshape the same component. Lots of fun!
The beauty of this particular design is of course all mine.
-
Hebrides, in what way are the meshes inaccurate? I was under the impression every edge was just a vector, and so by joining vectors you made faces. I cant see any inaccuracy in this, although apparently im missing something.
-
oooh! my head hurts now.
feel like I've just taken the red matrix pill and the real hole ridden nasty model behind the SU illusion facade has just been revealed.
-
Solo, I highly enjoyed your plant collection and admire these soft models you are developing, but I look forward even more to your tutorials so I can model my own custom designs. Having that ability would be more valuable to me than purchasing a package of someone else's designs. Plants and trees, on the other hand, are a must!
Thanks again for all your great work
Advertisement