For all of you that are interested in procedural generation and stuff, there will be an interview with Miguel Cepero (Procedural World Blog, VoxelFarm) on Sunday, Februaray 3rd on AIGameDev.com.
FYI - Procedural World Developer - Interview
- Thread starter Skaldarnar
- Start date
Okay, here are my messy notes on the interview. Maybe they are interesting for some of you. I tried to provide some useful links where I could find them 
His blog: http://procworld.blogspot.com/
----------------------------------------------------------------------------------------------------------------------------------------
Voxels/Polygons
- Dual Contouring
-> Voxel to Polygons
Terrain
-not much noise funcitons like Perlin (overused!).
- Heightmap (Perlin)
-> apply naturla filter (Erosion)
-> 3d noise for local detail (e.g. worley noise)
- texture splatting -> material information
- Hugues Hoppe's "Parallel controllable texture synthesis" for terrain?
Frankesterrain
-> Stitching terrain samples together
-> noise + pre-generated (use samples from simulations)
-> erosion/decay baked into sample data
-> Wang Tiles (blogpost)
Terrain Details
- little stones instead of grass in deserts/caves
- material based placement
- set of rules to determine tree/building placement
Trees/Forests
-> Cellular Automata (Calgary university - botany) for forests
Individual Trees
-> Space colonization (goal driven)
-> L-System do not understand what they are representing (out of control)
Caves
-> challenging/expensive
-> occlusion test (what is hidden)
-> generate dungeon system
-> corridors/rooms (disturbed lines)
-> but where to put a cave?
Buildings
-> exterior/facade
-> outside is quite simple
-> CityEngine
-> polygonal meshes as output
-> turn back to voxels
Biomes
-> Repetive Textures
-> Wang Tiles
Water
1. What is ocean, what is land
2. Wind directions to determine humidity
-> Wind simulation
3. River (place source in moutains, trace way to the ocean)
Tiles
-> Process tiles in isolation
-> Use tiles for biomes/sections/areas
Issues/Plans
- combination of coding/programming and artistic style
- use real world data to feed the system -> machine learning
Tips/Tricks
-> "Everyone who has done a little is already an expert!"
Links (from the chat)
- Wang Tiles
- Tile genetics
- Bachelor dissertattion - procedural planets
His blog: http://procworld.blogspot.com/
----------------------------------------------------------------------------------------------------------------------------------------
Voxels/Polygons
- Dual Contouring
-> Voxel to Polygons
Terrain
-not much noise funcitons like Perlin (overused!).
- Heightmap (Perlin)
-> apply naturla filter (Erosion)
-> 3d noise for local detail (e.g. worley noise)
- texture splatting -> material information
- Hugues Hoppe's "Parallel controllable texture synthesis" for terrain?
Frankesterrain
-> Stitching terrain samples together
-> noise + pre-generated (use samples from simulations)
-> erosion/decay baked into sample data
-> Wang Tiles (blogpost)
Terrain Details
- little stones instead of grass in deserts/caves
- material based placement
- set of rules to determine tree/building placement
Trees/Forests
-> Cellular Automata (Calgary university - botany) for forests
Individual Trees
-> Space colonization (goal driven)
-> L-System do not understand what they are representing (out of control)
Caves
-> challenging/expensive
-> occlusion test (what is hidden)
-> generate dungeon system
-> corridors/rooms (disturbed lines)
-> but where to put a cave?
Buildings
-> exterior/facade
-> outside is quite simple
-> CityEngine
-> polygonal meshes as output
-> turn back to voxels
Biomes
-> Repetive Textures
-> Wang Tiles
Water
1. What is ocean, what is land
2. Wind directions to determine humidity
-> Wind simulation
3. River (place source in moutains, trace way to the ocean)
Tiles
-> Process tiles in isolation
-> Use tiles for biomes/sections/areas
Issues/Plans
- combination of coding/programming and artistic style
- use real world data to feed the system -> machine learning
Tips/Tricks
-> "Everyone who has done a little is already an expert!"
Links (from the chat)
- Wang Tiles
- Tile genetics
- Bachelor dissertattion - procedural planets
My favorite part
Good show and the guy seems nice and happy to help discuss stuff. #gameai on Freenode might be a good channel to idle in to catch more nuggets of wisdom.
We might not have the sharp focus of any of the one-man closed source projects, but we're still in a good place IMHO
Good show and the guy seems nice and happy to help discuss stuff. #gameai on Freenode might be a good channel to idle in to catch more nuggets of wisdom.
yep, already added the channel to my pidgin list.
Thanks for the irc link, will hop in whenever i have time
So, here's the record of the #gameai chat debating rivers in (infinite) procedural worlds.
I was in there for that, yay! Saw the little doodle site too, handy.
I like the notion of limited length for rivers in infinite worlds based on height vs ocean. Start rivers up in mountains and let them run down the sides. If the river hasn't hit ocean after a certain length then potentially cancel it or create a lake. Could base it on drop over distance - if the river hasn't dropped appreciably over a few hundred meters then create a lake for it to empty into.
I like the notion of limited length for rivers in infinite worlds based on height vs ocean. Start rivers up in mountains and let them run down the sides. If the river hasn't hit ocean after a certain length then potentially cancel it or create a lake. Could base it on drop over distance - if the river hasn't dropped appreciably over a few hundred meters then create a lake for it to empty into.