tournamentdan

Yes. Frank Oz.

This is kind of what I felt like yesterday.

No just one cube. All of the solid interior walls on the first floor can be created from the bottom plane. Any doorway would need a two sided plane. The side planes can make up the inside of the outside walls. Along with the widows, doors, paintings.... And the top plane is the ceiling. Each bedroom would have one door and one closet door. So each bedroom add 8 triangles. Then a entry for kitchen, living room, laundry. 4 triangle each. Say a closet at the main door so you can hang your coat up. And maybe a pantry in the kitchen. Another 8 triangle. Two bath. 8 triangle. So a four bed room one story house could have as little as 82 triangles for a base (primitive) mesh. 72 for inside and 10 for out side. Maybe you did talk me into a combination of using both LOD. However, once the geometry shader gets going. You will not need to load the interior on the memory. Just create a primitive interior. Tessellate and boom. Edit: No your right. One cube plus 30 planes.

I was not thinking of interior doors. Any interior wall that has a door way would need its own plane. The rest can be brought up from the bottom plane. For every interior door add 4 triangles.

Why not? The interior walls can be created from the bottom plane of the inside cube. Also you could have the tile textures or carpet or hardwood floors. The left and right inside cube planes could make the inside section of the walls, windows, and doors. And even cracks in the paint. Same for the front and back planes. If you want to do multiple floors you could place a two sided plane at each floor level. One side will be the ceiling and the other interior walls for the second floor. You would not be able to create the furniture and other items.

Yes but think about it like this. My two cubes (inside/outside) both collectively have 22 triangles. Each cube has one set of textures. If you add up the triangles for three maybe four LOD and each LOD has their own texture. The LOD memory footprint will be much larger.

Here is a thought. When I first learned about tessellation I had an idea on how to create a tree from just a few planes. I have not had much time to try some of my new ideas. My idea involved what is called a two sided plane. A two sided plane is just two planes facing back to back almost in the same x,y,z. Just a tiny bit offset on one axis. So what if we use a cube with the normals on the outside. Then make a cube a little bit smaller and have the normals face the inside. Then you could tessellate the outside and inside when needed. The base mesh for each house or building would have eleven faces. I know your saying two cubes would be twelve faces. But you would not need the bottom face for the outside cube. I suppose if you would want to populate your game with a bunch of houses, buildings,trees, ect.. you still may want to use a billboard to help with the long camera shots. @Rick The concepts are the same. We can only render a certain amount of triangles. With both techniques all you are trying to do is render the most amount of triangles closest to the camera. To decrease and increase a models poly count with tessellation is LEVEL OF DETAIL.

The outside and the inside of a building can be different models . The inside of the house does not necessarily need to be there until the player gets close to the house. This does not mean you need a billboard of the building or three or more LOD of the house. You could have a cube with textures of the house. As the camera gets closer you add more detail. Then when you get real close, then you could draw the inside of the house. Or maybe create all of the inside walls with a couple planes. Think about what a geometry shader could do here. When the player gets close. You could create the entire inside of the house and all of the models(kitchen cabinets, soda cans, tv, couch etc...) in real time. With out all of those models loaded on memory. How about infinite terrain? All can be done with geometry shader and tessellation.

Don't you worry about me A-hole. When I am ready. I will gladly show what I have been working on. Currently I have a lot on my plate. Rebuilding a house, running my business, developing a cnc for friend of mine. Oh and currently today I have been busy correcting a nincompoop spreading lies about things they obviously know nothing about.

You must of missed the part where I said. No. Look at the top of the screen. The subject reads: LOD system questions.

I agree with every thing you said except it may be a little harder to tessellate an individual face or triangle. The only way I can think to tessellate a specific triangle is if your model had vertex data(or vertex group) for each individual vertex. That might get a little larger for file size. Not sure, but that would be something to look into. Or if you want to tessellate a certain section of a model. You could create more seams for texture and tessellate that one texture. Or do a raycast and place a decal and tessellate the decal only.

I am glad you brought up Uncharted. Since I have had conversations with the artists that made the characters for Uncharted. The picture that you showed is the only version of that character since it is a third person game. The character is always at the same distance from camera and does not need LOD. And it is the same model they used for cut scenes and promotional media. But lets just say it was a first person. And that is the highest amount of detail you want. Which I agree with you. That is a good amount of detail. But, with tessellation instead of having three or four LOD to get up to that point. You could have one low poly model for the far distances. And if that same low poly model gets closer to the camera. You can tessellate it to the same poly count as your posted picture. So you see tessellation is a level of detail system In its own.

I am glad you made this comment. It shows that you make up **** so it sounds like you know what you are talking about. Since GTAV has come out I have been very addicted to it and know a lot about it. Your right. Gtav does not use tesselation. Let's dive into why... Rock star first started to develop gtav, five years ago for hardware that was released to the public about seven years ago. And the technology for those consoles were developed about nine years ago. The ps3 does not support dx11 or opengl4. There were earlier version of tesselation before dx11 and opengl4. But they were not too performace friendly. Tesselation became realtime performance friendly after dx11 and opengl4. Which is not possible in the old consoles. Because GTAV is so large and uses so many LOD. It takes about five minutes and more for online to load the game. If the old consoles could of supported real time tesselation. The games would load a lot faster or instead of having all of those lods and extra textures.You could have even more model assets in game with the same amount of load time and memory. The next big thing will be with the geometry shader and the ability to both create and delete a vertex on the fly.

People, stop assuming on what you do not know. You are only confusing yourself and other people. Tesselation can do exactly what you explained. Your base mesh for a face can be 100 poly. As the face gets closer you can tesselate it to around 1000. The displacement map puts the vertex where they need to be. There is no other calculation.

My god Yougrove. It is clear you do not know what you are talking about. You need to learn more about it before you comment. You do not need to divide the triangles into an ungodly amount. You can tesselate sections of a model. You can tesselate a cube into any model you want. Tree, car, barrel, a cute character model of yougroove of which I could run over repeatedly with my newly tesselated car.

The placement of vertice depends on the displacement map. The quality of the displacement map depends on the artist. When making textures for game assets. You need to create a high poly model. Then a low poly model. I can now bake textures from the high to the low. Then use the low poly for a (primitive) in a game. And adjust the level to my liking in game real time. It doesn't make detail magically. The artist has already made the detail. Which is contained in the displacement map. Tesselation does not curve the mesh. It only divides each triangle of a model or each triangle in a decal.

What is so complicated about a displacement map and choosing a number to divide the triangles by based on the distance from the camera. This has not been widely used because the general public has not had the hardware. Game studios are not going to make a game for a couple people. As the last couple years have gone by. More and more people have up graded their hardware. So now it is becoming more and more popular.This makes game asset creation easier and it also clears up a lot of memory. Which would make a game run better and faster or, would allow you to put a bunch more **** in a game.

I guess this just falls in the category of old dogs and new tricks. I understand that the same technique has been used for the last twenty years. But with new technology comes new techniques. And also with new technology come hesitation, miss-understanding, and miss-information. Tesselation is not only for the real close up. It will not allow you to render more triangles. It will however allow you to render more triangles closer to the camera. You do not have to go from six polygons to a million on a model. You can divide by smaller numbers than ten. As for performance. The Gtx-480 can tesselate over 1.6 billion triangles per second. Not to shabby. Since I must be having a hard time explaining it. Read the following link. It is a good general info on tesselation in games. http://www.nvidia.com/object/tessellation.html

There are different levels off tesselation. You would not want to tesselate an object for every inch the camera moves. This would be costly. I have not seen how josh has implemented it but the user needs to determine which level of tesselation and how far. @yougroove. The second picture of the dragon is obviously over tesselated for a video game. Instead of having three or four lod's of a barrel. You can have one primitive barrel. Just for the neck of it lets say that barrel has six polygons. You could then tesselate it up to 460. Then if the camera gets closer, then tesselate it to 660. You could even use a decal on a section of the model and then only tesselate that decal part. Think of how much more memory you will have when you do not have all those lod loaded and just hanging around. Waiting to get used.

Because you can create one low poly model. Then tesselate that model at the the same distances that you would change the lod .Your mind will really be blown when josh gets the geometry shader going. Then you will not have to load certain models at start up.

Not every game gets the same amount of fps. When you first start your game your world is small. As your game grows the fps come down. The final target range should be around 60 fps. So checking distance from an object to the camera every 200 frames will not do. As your games games evolve your code for this will have to also. Twice a second should be good enough to fool the eye. So if your game is getting 60 fps. Then every 30 fps is fine. Also there is no reason to disable texture maps at long distances because the low quality lod's should not have specular, normal, etc... Low quality lod should have only one low res texture. Also less bones. Having said all of this. You do not need lod for opengl4. So I am guessing you need it for the low end renderer.

Do you want to use lod for the fall back renderer or opengl4 ?

Looks great. Just curious. What is the poly count after tesselation? And what is your frame rate in your test? Have you tried tesselating a decal yet?

The version sold on steam is 3.1. Same as the preorder version here. The only difference between the steam (inde) and the preorder version is the steam (inde) version can only be coded in Lua. Instead of c++ and Lua.

Are you saying the base priced should never go up? No matter how many features are added the base price should be the same? I do not understand this thinking. If I go buy a new car and add some options (features) the cars price will go up. So how can you expect the base price for LE to stay the same forever, no matter how many features are added. And the only people that absorb the cost and profits for creating that new feature are the return customers.