First of all I have to go out in the nature, or somewhere and take mmaaanny picture of the same object. This sound easy, but if you want no stretched textures and a good topology it becomes more difficult.
In the second part I have to bring all the stuff on my Computer. Sometimes it is necessary to mask out every picture. Luckily I will build a little studio in the next year which will automate this work for (just for small objects I can take home).
The third step is to compute the orientation of each picture to the other pictures, then create a dens point cloud (many points in a space). After this I can create a first mesh from the
points and, if I want so a texture based on the taken pictures as well.
I close meshes by hand in Blender, like the ancient pillars or some of the rocks.
Next I export the mesh and the textures to Blender
I decimate the mesh using Meshlab and then transfer the textures to the mid- and low-poly versions via Blender texture-baker.
Of coarse I have to create Uvs for each new model.
The last step is to export all Textures ( diffuse / albedo, normal-map, and displacement-map ), create a second, more low resolution texture set and
upload all textures and different file types of the meshes to CGTrader.
Sounds Interesting? have a look at CGTrader!
I started to create Textures from scanned surfaces. The pro is that the result looks extremely realistic. The con is that I have to do much more work. I already uploaded some texture-packs to CGTrader, ready to be used in games and offline render projects.
I started capturing meshes without holes via custom masks, the outcome is a fully scanned mesh. Next step will be to create a lightning condition which allows me to capture de-lighted pictures
for best possible meshes.
visit the links to watch the models:
Highres. Pumpkin Seed Bread
Dark Bread 02