@ Snefer- thanks man, we love your latest Grime district stuff!
Your specs are signifying that you're getting a terrible framerate. The grass can't possibly have so many polys to be doing that, it's very likely a drawcall issue. I know that Unreal has a system of combining objects together to cut down drastically on drawcalls. I don't know if it's practical for grass, but it's worth looking into.
Perhaps adjusting the color of the grass to match the terrain, and setting up a distance fade-out could do the trick too.- Makkon
Hey quit snooping on our perf! jk, but yeah we realized that painting the individual vegitation clumps was creating way too many draw calls so we started exporting larger clumps with the fit to terrain option checked. It drastically improved the frame rate but tweaks out if we paint the grass on any surface that has too much of a slope change.
Here are a couple shots of our Oak tree that Danny created.
Also here is a link to one of our tutorials for creating a dirt path using 3ds Max and CryEngine3's Spline system.
did you find a way to blend your overhangs?
I need to make something similar and seamlessly blending geometry with the tessellated height map seems near impossible at the moment : [
..yeah we realized that painting the individual vegetation clumps was creating way too many draw calls so we started exporting larger clumps with the fit to terrain option checked. It drastically improved the frame rate but tweaks out if we paint the grass on any surface that has too much of a slope change.
Well that sucks donkey balls. Does the geometry instancing feature work with the current build to help keep the drawcalls low?
did you find a way to blend your overhangs?
I need to make something similar and seamlessly blending geometry with the tessellated height map seems near impossible at the moment : [
Here are a couple updates to the overhang and although I'm not perfectly happy with the solution I think it will work for now.
This first shot shows the lod1 state of the overhang props and these are essentially the excact same geo except without prop grass.
Here is a close up of that same area with the prop grass loaded in and also the painted vegetation rocks and bushes. I made a lopped off version of the overhang to get the scallop effect I had in my paint-over.
This is the top down view of that same area so you can see the prop\terrain line. Its already pretty subtle here.
and here is a close up of the terrain prop boarder with all the vegetation loaded to help mask it.
So I think it's basically... use the same shader values & textures, keep the surfaces near-parallel at the join, then cover with detail meshes. Is this correct? A good workflow, and works in just about any engine.
So I think it's basically... use the same shader values & textures, keep the surfaces near-parallel at the join, then cover with detail meshes. Is this correct? A good workflow, and works in just about any engine.
I think the seam is here?
Correct, for these rocks the diffuse textures are pretty much desaturated and then I'm using vertcolor to match the color of the terrain. I usually sample the color directly from the megaTerrain Diffuse for a close starting point.
Your'e right, that is the correct seam. My seam line was a decoy to keep you on your toes
wow i'm totally following this. Looks gorgeous so far! I always wanted to create massive environments, but so scared too. Can totally learn a lot from this.
Both the videos so far are great, thanks for sharing! Any news on what videos we can look forward to next? (I just can't get enough of your guys' work)
Well that sucks donkey balls. Does the geometry instancing feature work with the current build to help keep the drawcalls low?
Geometry instancing reducing memory usage not the amount of draw calls.
I have an more practical question.
Did you generated entire terrain in WM, or you scultped it by hand and then processed in WM ?
If you generated entire thing in WM, how would you get scale right for gameplay purposes ?
I personally found that WM is all and nice, but when it comes to something more practical that fancy screenshots (like actually playable level) it's pretty much useless on it's own.
And it's still best thing to do is to sculpt terrain in your editor and then maybe export it to WM for some erosion filters.
Geometry instancing reducing memory usage not the amount of draw calls.
I have an more practical question.
Did you generated entire terrain in WM, or you scultped it by hand and then processed in WM ?
If you generated entire thing in WM, how would you get scale right for gameplay purposes ?
I personally found that WM is all and nice, but when it comes to something more practical that fancy screenshots (like actually playable level) it's pretty much useless on it's own.
And it's still best thing to do is to sculpt terrain in your editor and then maybe export it to WM for some erosion filters.
This level started out entirely in world machine and then I exaggerated some of the ravines in Photoshop to get the layout I wanted.
WorldMachine definitely changes up the pipeline but I think It's well worth it. For instance, creating a map the same size as this test map for GuildWars 2 would take us months just to paint the terrain and blend all the textures. Now you can get far better results in a half day.
Even if you have to totally wipe your layout its still far quicker than a standard terrain workflow.
Also you can still use all your regular terrain sculpting tools in the editor so if designers want to totally flatten out an area or remove a mountain they can just block it in quickly, then the artist can export the height map and run it through world machine again.
This level started out entirely in world machine and then I exaggerated some of the ravines in Photoshop to get the layout I wanted.
WorldMachine definitely changes up the pipeline but I think It's well worth it. For instance, creating a map the same size as this test map for GuildWars 2 would take us months just to paint the terrain and blend all the textures. Now you can get far better results in a half day.
Even if you have to totally wipe your layout its still far quicker than a standard terrain workflow.
Also you can still use all your regular terrain sculpting tools in the editor so if designers want to totally flatten out an area or remove a mountain they can just block it in quickly, then the artist can export the height map and run it through world machine again.
Thanks.
But Here come another question. Don't you think it's more issue of tools, rather than pipeline itself ?
With alpha masked brushes and nice setup of generated alphas (yeah in World Machine) and drag tool, you can quickly sculpt anything you want. And designer can still have fine control over scale and where players should go (or shouldn't).
I ask because I have exactly opposite experience. I a;ways generated nice terrain in WM, that was looking nice for screenshots, but it never fit for gameplay on it's own and I usually had to wipe out pretty much everything and sculpt by hand to establish layout I need, and then export to WM for erosion, and then make changes in editor.
Of course Sandbox for some odd reason doesn't have Alpha brushes, but that's another matter (;.
But Here come another question. Don't you think it's more issue of tools, rather than pipeline itself ?
With alpha masked brushes and nice setup of generated alphas (yeah in World Machine) and drag tool, you can quickly sculpt anything you want. And designer can still have fine control over scale and where players should go (or shouldn't).
I ask because I have exactly opposite experience. I a;ways generated nice terrain in WM, that was looking nice for screenshots, but it never fit for gameplay on it's own and I usually had to wipe out pretty much everything and sculpt by hand to establish layout I need, and then export to WM for erosion, and then make changes in editor.
Of course Sandbox for some odd reason doesn't have Alpha brushes, but that's another matter (;.
I think the real value is getting the diffuse texture and flow maps to match the terrain features automatically.
In my opinion BattleField 3 has some of the best looking terrain systems for a released game and I think they use World Machine quite a bit. I would be curious to learn exactly when they introduce the WorldMachine features. My guess is they talk with the designers and get a general idea which ares are battle zones and then they go wild with the vistas boardering them.
Now what I think would be ideal is an evolution of Z-brushe's insertMulti Mesh brushes were they also keep uv information. If a system like that could be combined into a voxel terrain editor I would be in heaven.
In my opinion BattleField 3 has some of the best looking terrain systems for a released game and I think they use World Machine quite a bit. I would be curious to learn exactly when they introduce the WorldMachine features. My guess is they talk with the designers and get a general idea which ares are battle zones and then they go wild with the vistas boardering them.
@ Santewi- Ha I thought I read that somewhere, thanks for posting those links. I would love to get my hands on some Frostbite 2.
As I mentioned before one of the goals with this project is to have genre variety and so far the level has been something you might see in a high fantasy universe similar to Guildwars2 or Skyrim. These next additions are still early wips of blocking in a monolithic sci fi tower that reaches through the atmosphere and up into space (aka space elevator)
Here are some initial sketches to brainstorm what kinds of modular pieces I will need.
This shot shows some blocked in grey box geometry and at this stage I'm just playing around with scale and shape intersections.
Heres a quick paint-over of the previous shot along with some supporting reference that I would like to add in.
Some of the paintover elements blocked in and also start adding placeholder materials like glass, concrete, metal etc.
The same shot exploded out so that you can see some of the placeholder shapes. The idea is to try and keep the modular pieces down to a minimum so that I can go back and polish each one once I'm happy with the far read of the tower.
I far shot of the tower.
Mid level look.
Looking up from the mountain base.
another quick paintover with some ideas on where I'd like to take it.
far away shot with the tower scaled down so it doesnt dwarf the mountain as much.
paintover to see what it would look like if I attach it to a space station.
At this stage I'm pretty happy with the silhouette of the tower so I will start creating final assets from the placeholder pieces.
Looking sexual ! you better use some of those volumetric clouds on that tower to add more eye candy. Keep it up, very inspiring !
I was thinking about something along these lines but with a realistic style applied to it.
Looking sexual ! you better use some of those volumetric clouds on that tower to add more eye candy. Keep it up, very inspiring !
I was thinking about something along these lines but with a realistic style applied to it.
Thanks man, yeah I've messed around with some of the 3D Clouds in cry but every time I start copying them around to build up the atmosphere things start tweaking out. Any time I add more than 1 cloud entity they start flickering on and off. Also if I make them too large the sky turns black briefly as I fly closer to the clouds.
Thanks man, yeah I've messed around with some of the 3D Clouds in cry but every time I start copying them around to build up the atmosphere things start tweaking out. Any time I add more than 1 cloud entity they start flickering on and off. Also if I make them too large the sky turns black briefly as I fly closer to the clouds.
Have you seen any of these errors before?
There seem to be an issue with the 3.4 build. Volumetric clouds are buggy and you can't use the cloud shadows stencil. I think that geometry instancing is also broken but I'm not sure if its only limited to DX9 users.
I stick to using the 3.3.9 build which in my opinion is the most functional one at the moment.
Unless you're planning to use some DX11 features (tessellation etc...) transferring over to 3.3.9 should be an easy task. You'll have to rebuild your TOD tho.
The next build after 3.4 should be fixing all those issues according to crytek but the release date is tbd.
They said on blog post that they were hand sclupting terrain first. Exporting heightmaps to WM for erosion and mask generation, importing heightmap bacl to editor and tuning it again.
A process I essentially described above. But it relies on good in-editor tools in first place. On other hand you can quickly see what's wrong and what's not.
Switched the level over and now the clouds are working. Thanks for the heads up!
One problem that came up is that when the trees switch to sprites they are washed out and very bright in the distance. In the past when this would happen, regenerating the terrain texture would fix it but now that doesn't work.
I had the same problem before where sprites still showed up. You could go to vegetation, click on the object you want to fix and just uncheck 'Use Sprites.'
hmm i'm not sure about cryengine as I have very little experience in it. If it were UDK I would try disabling lighting on the sprite sheet, and rendering it out with a controlled emmisive. Alternatively the fog tends to illuminate things, It might be worth seeing if there are any values you can tweak in the fog settings to reduce the ammount that it 'lights the surroundings. Hope this helps and look forward to seeing more =D
@ Bonkahe- Thanks, once we launch the actually class we will release the level as an example.
@ e-freak- the trees are using Cryengines automated vegitation sprite system. They worked fine in the 3.4 build so I'm gonna stick with that until I find a solution.
@ Lo Pan- It was good seeing you too Dan. Good luck on the Arena test.
@ RuthK- I would do that but my computer would melt since I have a million trees
@ AzzaMat- yeah I'll look at the fog settings and see if theres anything that relates to the sprites. Thanks.
Here are a couple more screens of the firTrees Danny and I made.
Here are a couple shots looking at the horizon from the base of the tower. I used world machine to make a few giant props that help hide the square boarder of the map.
Also here is a link to some intro to 3dsMax videos we made.
@samcole,Eric Chadwick,Computron- they will remain streaming for right now. Downloadable options might be available to students once the class launches.
@inside- Thanks, I'll try that and see if they light better. Ideally I'd like to reserve hand doing the sprite lods as a last resort but I should try it anyway and see if perf is better or worse.
Here are a couple shots of one of the floors on the tower along with a time lapse paintover to show the iterative process.
Replies
@ Fingus- lol, best quote yet!
@ Snefer- thanks man, we love your latest Grime district stuff!
Hey quit snooping on our perf! jk, but yeah we realized that painting the individual vegitation clumps was creating way too many draw calls so we started exporting larger clumps with the fit to terrain option checked. It drastically improved the frame rate but tweaks out if we paint the grass on any surface that has too much of a slope change.
Here are a couple shots of our Oak tree that Danny created.
Also here is a link to one of our tutorials for creating a dirt path using 3ds Max and CryEngine3's Spline system.
http://link.brightcove.com/services/player/bcpid1783827065001?bckey=AQ~~,AAAAv1s33Ak~,N2Teo4GAFCq8U4OgL4UdHC2Nc4O95Ske
I need to make something similar and seamlessly blending geometry with the tessellated height map seems near impossible at the moment : [
Here are a couple updates to the overhang and although I'm not perfectly happy with the solution I think it will work for now.
This first shot shows the lod1 state of the overhang props and these are essentially the excact same geo except without prop grass.
Here is a close up of that same area with the prop grass loaded in and also the painted vegetation rocks and bushes. I made a lopped off version of the overhang to get the scallop effect I had in my paint-over.
This is the top down view of that same area so you can see the prop\terrain line. Its already pretty subtle here.
and here is a close up of the terrain prop boarder with all the vegetation loaded to help mask it.
Hope that helps
I think the seam is here?
Correct, for these rocks the diffuse textures are pretty much desaturated and then I'm using vertcolor to match the color of the terrain. I usually sample the color directly from the megaTerrain Diffuse for a close starting point.
Your'e right, that is the correct seam. My seam line was a decoy to keep you on your toes
Geometry instancing reducing memory usage not the amount of draw calls.
I have an more practical question.
Did you generated entire terrain in WM, or you scultped it by hand and then processed in WM ?
If you generated entire thing in WM, how would you get scale right for gameplay purposes ?
I personally found that WM is all and nice, but when it comes to something more practical that fancy screenshots (like actually playable level) it's pretty much useless on it's own.
And it's still best thing to do is to sculpt terrain in your editor and then maybe export it to WM for some erosion filters.
This level started out entirely in world machine and then I exaggerated some of the ravines in Photoshop to get the layout I wanted.
WorldMachine definitely changes up the pipeline but I think It's well worth it. For instance, creating a map the same size as this test map for GuildWars 2 would take us months just to paint the terrain and blend all the textures. Now you can get far better results in a half day.
Even if you have to totally wipe your layout its still far quicker than a standard terrain workflow.
Also you can still use all your regular terrain sculpting tools in the editor so if designers want to totally flatten out an area or remove a mountain they can just block it in quickly, then the artist can export the height map and run it through world machine again.
But Here come another question. Don't you think it's more issue of tools, rather than pipeline itself ?
With alpha masked brushes and nice setup of generated alphas (yeah in World Machine) and drag tool, you can quickly sculpt anything you want. And designer can still have fine control over scale and where players should go (or shouldn't).
I ask because I have exactly opposite experience. I a;ways generated nice terrain in WM, that was looking nice for screenshots, but it never fit for gameplay on it's own and I usually had to wipe out pretty much everything and sculpt by hand to establish layout I need, and then export to WM for erosion, and then make changes in editor.
Of course Sandbox for some odd reason doesn't have Alpha brushes, but that's another matter (;.
I think the real value is getting the diffuse texture and flow maps to match the terrain features automatically.
In my opinion BattleField 3 has some of the best looking terrain systems for a released game and I think they use World Machine quite a bit. I would be curious to learn exactly when they introduce the WorldMachine features. My guess is they talk with the designers and get a general idea which ares are battle zones and then they go wild with the vistas boardering them.
Now what I think would be ideal is an evolution of Z-brushe's insertMulti Mesh brushes were they also keep uv information. If a system like that could be combined into a voxel terrain editor I would be in heaven.
I'll email Obama and see what he thinks
http://dice.se/publications/terrain-in-battlefield-3-a-modern-complete-and-scalable-system/
http://blogs.battlefield.com/2012/08/inside-dice-alborz/
@ Santewi- Ha I thought I read that somewhere, thanks for posting those links. I would love to get my hands on some Frostbite 2.
As I mentioned before one of the goals with this project is to have genre variety and so far the level has been something you might see in a high fantasy universe similar to Guildwars2 or Skyrim. These next additions are still early wips of blocking in a monolithic sci fi tower that reaches through the atmosphere and up into space (aka space elevator)
Here are some initial sketches to brainstorm what kinds of modular pieces I will need.
This shot shows some blocked in grey box geometry and at this stage I'm just playing around with scale and shape intersections.
Heres a quick paint-over of the previous shot along with some supporting reference that I would like to add in.
Some of the paintover elements blocked in and also start adding placeholder materials like glass, concrete, metal etc.
The same shot exploded out so that you can see some of the placeholder shapes. The idea is to try and keep the modular pieces down to a minimum so that I can go back and polish each one once I'm happy with the far read of the tower.
I far shot of the tower.
Mid level look.
Looking up from the mountain base.
another quick paintover with some ideas on where I'd like to take it.
far away shot with the tower scaled down so it doesnt dwarf the mountain as much.
paintover to see what it would look like if I attach it to a space station.
At this stage I'm pretty happy with the silhouette of the tower so I will start creating final assets from the placeholder pieces.
I was thinking about something along these lines but with a realistic style applied to it.
Thanks man, yeah I've messed around with some of the 3D Clouds in cry but every time I start copying them around to build up the atmosphere things start tweaking out. Any time I add more than 1 cloud entity they start flickering on and off. Also if I make them too large the sky turns black briefly as I fly closer to the clouds.
Have you seen any of these errors before?
There seem to be an issue with the 3.4 build. Volumetric clouds are buggy and you can't use the cloud shadows stencil. I think that geometry instancing is also broken but I'm not sure if its only limited to DX9 users.
I stick to using the 3.3.9 build which in my opinion is the most functional one at the moment.
Unless you're planning to use some DX11 features (tessellation etc...) transferring over to 3.3.9 should be an easy task. You'll have to rebuild your TOD tho.
The next build after 3.4 should be fixing all those issues according to crytek but the release date is tbd.
A process I essentially described above. But it relies on good in-editor tools in first place. On other hand you can quickly see what's wrong and what's not.
Switched the level over and now the clouds are working. Thanks for the heads up!
One problem that came up is that when the trees switch to sprites they are washed out and very bright in the distance. In the past when this would happen, regenerating the terrain texture would fix it but now that doesn't work.
Has anyone ran into this?
Stellar work man, the tower is awesome :thumbup: Will be following your progress with interest
Was great to catch up with you in Seattle and check out the new studio
Arenanet moved back the deadline for character submissions so I was able to send them in a model
Dan
@ e-freak- the trees are using Cryengines automated vegitation sprite system. They worked fine in the 3.4 build so I'm gonna stick with that until I find a solution.
@ Lo Pan- It was good seeing you too Dan. Good luck on the Arena test.
@ RuthK- I would do that but my computer would melt since I have a million trees
@ AzzaMat- yeah I'll look at the fog settings and see if theres anything that relates to the sprites. Thanks.
Here are a couple more screens of the firTrees Danny and I made.
Here are a couple shots looking at the horizon from the base of the tower. I used world machine to make a few giant props that help hide the square boarder of the map.
Also here is a link to some intro to 3dsMax videos we made.
http://link.brightcove.com/services/player/bcpid1756048041001?bckey=AQ~~,AAAAv1s33Ak~,N2Teo4GAFCqLdBwbw7r6pAoW85WLHZlM
Hm. Did you considered doing lod level with sprites by hand ?
@inside- Thanks, I'll try that and see if they light better. Ideally I'd like to reserve hand doing the sprite lods as a last resort but I should try it anyway and see if perf is better or worse.
Here are a couple shots of one of the floors on the tower along with a time lapse paintover to show the iterative process.
[ame="
A priority breakdown of assets that I will need to create first
me when I saw CeleryLand.
the new tower paintover is sick! Can't wait to see how that develops.