This is the second time something has mentioned the mountain/rock in the background. I guess I'm seeing it the way I had it planned, but it seems others aren't seeing what I'm seeing obviously. I'll try to explain what I mean and I'll take notes for changes to make for my next update.
Basically, the rocky stone/mountain you see is there to hide the horizon line, making my task much, MUUCH easier. The mountain is also, not technically a mountain. I think you are mistaking the scale, which is due to my poor execution. The "mountain" is about... 30-40 meters away from the temple itself. The scene is located in a small box canyon, where the water has pooled up to form a pond. That small tree is not small because of the distance, it is just small because I wanted a small tree there. I'll fix that and put two similarly-sized trees to better gauge the distance.
Hope that cleared some things up a bit. I'll make it clearer next time. Thanks for pointing that out!
This is what I get for trying to lurk on Polycount during class when I should be paying attention. I completely miss things that were mentioned earlier, aha.
That totally makes more sense now. I'm totally going to stalk this project for updates like you wouldn't believe.
Ohh I didn't mean someone else in this thread noticed it! I just meant that another friend of mine noticed it too. It may make sense, but I need my art to be clear just by looking at it the first time! So thanks!
But I know what you mean, I'm constantly lurking around here instead of listening in class. Believe it or not, sometimes I feel I am learning more from here than I learn in class! (Pretty rare, but it happens).
@Sasochicken: Yeah I agree that they lack contrast. I changed the leaf type due to mistake on my part. I was using a Canadian Maple leaf instead of the japanese maple leaf. I also decided to make my scene late summer to early autumn. Therefore the leaves are red. I also increased the contrast.
I'm not getting the contrast I'd like to have, I'll see what I can do to make the leaves seem less flat.
ouch rendering that...
all that overlapping alpha, try combining some leaves into large medium and small branch sets mapped onto a curving plane its not the polycount that will kill you here its the overdraw, you will be able to minmise and still acheive the same visual result with this techniques
I did a quick paint over for the leaves. For low poly trees, its all in the texture work. I think all you would need to do to get some variation in your leaves is make two leaf textures, one being darker and the other lighter so you can layer them on top of each other to get more lush/volume to your foliage. Also adding yellow/oranges to the tips of the leaves will break it up a bit and be more convincing. Oh and.. ADD SATURATION! Over saturate and then pull it back a bit till you're happy with your result
@ SasoChiken: Oh, I definitely understood what you meant. What I didn't quite understand was Shepheiro's statement. It's just a little unclear to me. I get an idea of what he's saying, but I don't quite get it.
@Saiainoshi Thanks for that paintover! I was just about to add a second leaf variation. I'll see what I can do without steering this into a fantasy tree. Need to keep it realistic.
ok not the best example, different style tree from a game i worked on a couple of years ago, Leon warren my lead at the time actually made the tree
you can afford alot more planes than these have, maybe 3 plane for every one that i have as your not rendering a forest, make the main plane cover and provide thickness like in the shot, then add a couple of other planes that are sparser to give a better silohette, but each plane should contain leaves and a branch so that you dont get magical floating leaves like you have cuttently
ahh I see where you're getting at. I'll take that into consideration. I need to redo the leaf texture since it is wasting UV space and I can get less overdraw. But I don't think the final rendering times will be much of an issue, since I'll have access to a render farm. But you are right, the quicker the renders the better, especially if I can maintain the same visual quality.
Mini update! Got a new tree made! Will be making some variants too. But I don't have time to bring them into photoshop and use them in my paintover. I'll get to that soon enough though (Note that I have terrible lighting going on at the moment. I just setup something quick to get a somewhat decent look)
I think you should try optimizing your Japanese maple so it's more appropriate for real-time applications though. One thing that could help you with making a diffuse is arranging the leaves from a top view, baking them for a normal and the alpha, and then use that.
That Japanese maple looks like it's around 2000 tris or more. While that would be okay if you only placed a couple in a scene, it's just better practice to maintain that high poly look while optimizing it further.
By the way, I know what you are talking about when you were referring to overdraw. Overdraw isn't really applicable to real-time game scenes, since alpha's are rendered using much simpler techniques than in a rendering program. Overdraw isn't something you need to worry about at all for games - it only is an issue when rendering scenes using raytracing.
By the way, I know what you are talking about when you were referring to overdraw. Overdraw isn't really applicable to real-time game scenes, since alpha's are rendered using much simpler techniques than in a rendering program. Overdraw isn't something you need to worry about at all for games - it only is an issue when rendering scenes using raytracing.
wrong wrong wrong...
alpha and overdraw is the most expensive things in all the realtime rendering pipelines ive worked with
alpha means pixels need to be drawn multiple times, one for each alpha plane passed through in that pixels line of sight.... 100 alpha planes in a row and your doing 100 times the work for each pixel....
this is why you will see face-on branch planes in alot of games as it gives the best form for the least overdraw per tree
I avoided that with my trees thankfully, even though I didn't know it was that important. I guess that's why I've gotten such good performance with them even with a thousand or more trees on screen at once.
I use an engine which uses alpha testing though, ( Cryengine2 ) which is a lot cheaper performance wise than alpha blending. It really depends on the engine with how the model should be made, because of the variety of ways shaders can be optimized and what not.
Additionally, vegetation textures in Crysis all utilize full mip-maps, so it isn't the same resolution of textures being drawn; thus less pixels are actually being traced on the alpha maps for trees as they transition mip-maps further away. Crysis's CryTif compression plugin for Photoshop also allows to save in all needed DXT formats and 3Dc.
Now, you may say that Cryengine2 is a demanding engine altogether. While it is, it was just poorly optimized. Cryengine3 is running on consoles, at the same level of quality that Crysis ran on PC with High settings applied, aside from the textures.
mipmaps- i dont know any full 3d engine that doesnt use mipmaps these days, not having it would kill the fillrate with modern texel density
alpha testing or stencil/cutout shaders help alot but there is still an overdraw cost compared to a normal solid shader due the unused Transparent sections of the mesh (always good to minimise these)
anyway dont want to hijack too much this is an interesting read
Replies
Basically, the rocky stone/mountain you see is there to hide the horizon line, making my task much, MUUCH easier. The mountain is also, not technically a mountain. I think you are mistaking the scale, which is due to my poor execution. The "mountain" is about... 30-40 meters away from the temple itself. The scene is located in a small box canyon, where the water has pooled up to form a pond. That small tree is not small because of the distance, it is just small because I wanted a small tree there. I'll fix that and put two similarly-sized trees to better gauge the distance.
Hope that cleared some things up a bit. I'll make it clearer next time. Thanks for pointing that out!
That totally makes more sense now. I'm totally going to stalk this project for updates like you wouldn't believe.
But I know what you mean, I'm constantly lurking around here instead of listening in class. Believe it or not, sometimes I feel I am learning more from here than I learn in class! (Pretty rare, but it happens).
And here is the wireframe:
I'm not getting the contrast I'd like to have, I'll see what I can do to make the leaves seem less flat.
@Jojones: No problem!
all that overlapping alpha, try combining some leaves into large medium and small branch sets mapped onto a curving plane its not the polycount that will kill you here its the overdraw, you will be able to minmise and still acheive the same visual result with this techniques
@Saiainoshi Thanks for that paintover! I was just about to add a second leaf variation. I'll see what I can do without steering this into a fantasy tree. Need to keep it realistic.
ok not the best example, different style tree from a game i worked on a couple of years ago, Leon warren my lead at the time actually made the tree
you can afford alot more planes than these have, maybe 3 plane for every one that i have as your not rendering a forest, make the main plane cover and provide thickness like in the shot, then add a couple of other planes that are sparser to give a better silohette, but each plane should contain leaves and a branch so that you dont get magical floating leaves like you have cuttently
http://i30.photobucket.com/albums/c346/Shepeiro/caspian5.jpg
I think you should try optimizing your Japanese maple so it's more appropriate for real-time applications though. One thing that could help you with making a diffuse is arranging the leaves from a top view, baking them for a normal and the alpha, and then use that.
That Japanese maple looks like it's around 2000 tris or more. While that would be okay if you only placed a couple in a scene, it's just better practice to maintain that high poly look while optimizing it further.
By the way, I know what you are talking about when you were referring to overdraw. Overdraw isn't really applicable to real-time game scenes, since alpha's are rendered using much simpler techniques than in a rendering program. Overdraw isn't something you need to worry about at all for games - it only is an issue when rendering scenes using raytracing.
wrong wrong wrong...
alpha and overdraw is the most expensive things in all the realtime rendering pipelines ive worked with
alpha means pixels need to be drawn multiple times, one for each alpha plane passed through in that pixels line of sight.... 100 alpha planes in a row and your doing 100 times the work for each pixel....
this is why you will see face-on branch planes in alot of games as it gives the best form for the least overdraw per tree
http://boards.polycount.net/archive/index.php/t-70435.html
I avoided that with my trees thankfully, even though I didn't know it was that important. I guess that's why I've gotten such good performance with them even with a thousand or more trees on screen at once.
I use an engine which uses alpha testing though, ( Cryengine2 ) which is a lot cheaper performance wise than alpha blending. It really depends on the engine with how the model should be made, because of the variety of ways shaders can be optimized and what not.
Additionally, vegetation textures in Crysis all utilize full mip-maps, so it isn't the same resolution of textures being drawn; thus less pixels are actually being traced on the alpha maps for trees as they transition mip-maps further away. Crysis's CryTif compression plugin for Photoshop also allows to save in all needed DXT formats and 3Dc.
Now, you may say that Cryengine2 is a demanding engine altogether. While it is, it was just poorly optimized. Cryengine3 is running on consoles, at the same level of quality that Crysis ran on PC with High settings applied, aside from the textures.
alpha testing or stencil/cutout shaders help alot but there is still an overdraw cost compared to a normal solid shader due the unused Transparent sections of the mesh (always good to minimise these)
anyway dont want to hijack too much this is an interesting read
I'd like to [ame="
Shep, interesting read indeed.
Prophecies, looking at your new tree makes me think it'll do quite OK with overdraw. There's some in there, but you can't avoid that completely.