So this is how I understand it:

An artist models the thing in Zbrush or Mudbox. It now has 10+ million polygon. He exports it in Zbrush as a normal map + low poly model. (this works?) The model now has ~20-30k polygons?

Now the artist imports the low poly model into 3dsmax or Maya, applies the normal map somehow as well, so it will look in the modelling program just like it would ingame. Then he starts adding bones and then skins the model. Then animation, then export. As I understand it, somehow all the textures of a model get "baked" into a single texture which the engine will understand?

Do people still make textures in Photoshop and then UV unwrap the model by hand, then apply the textures? I never understood how someone can make the laid out texture of a face in Photoshop, and somehow he knows that it will look good on the model. Do they paint the textures in Zbrush? But Zbrush doesn't have all the photoshop tools.

What if I want to make a terrain, or an indoor level? Do I just model the whole thing in 3dsmax? What if the textures don't show up in the engine the way they do in 3dsmax?

  • 3
    \$\begingroup\$ I think 2-3k polys (or less!) for a character would be closer to the truth (this varies of course, a MMORPG will use 1-2k, a fighter "versus" could use more as there are only 2 players). For the 'whole level in 3dsMax', its not done (usually) that way, items (floors, lamps etc) are made in 3dsMax and then an inhouse Level Editor is used to create the levels, placing the spawnpoints, lights etc. (as that's the job for the level designer, not the graphic guy) \$\endgroup\$ – Valmond Jan 25 '12 at 9:42
  • \$\begingroup\$ I think for textures on a model, you can do quite a bit directly in the 3d program. This way you could paint a rough sketch of the features (eg. eyes, eyebrows, nose, mouth, ears etc.) and then use that rough sketch as a guideline in Photoshop (or similiar). \$\endgroup\$ – Holger Jan 25 '12 at 13:46
  • 1
    \$\begingroup\$ As posed, this is subjective an unanswerable (except to say that almost every AAA workflow is different). What is the specific problem you are having? \$\endgroup\$ – user1430 Jan 25 '12 at 16:12
  • 1
    \$\begingroup\$ I am going to have to counter that downvote. Even though it may be a bit subjective the quality is good otherwise: which is worth encouraging. \$\endgroup\$ – Jonathan Dickinson Jan 25 '12 at 21:27
  • \$\begingroup\$ "I think 2-3k polys (or less!) for a character would be closer to the truth" That hasn't been true since the PS2 era. Character models nowadays break 10K easily. \$\endgroup\$ – Nicol Bolas Jan 25 '12 at 22:05

This is extremely difficult to answer because trade secrets such as the artists' pipelines for major developers are protected by long, strict NDAs. I'll do my best to summarize what I have learned as a student in the field.

The best answer I can start with is that it varies for every developer. What you described with the model first being created in Zbrush or Mudbox is fairly common, as I understand. The texture 'baking' is more varying. This is entirely dependent on what the engine requires. If it can handle a single texture containing all the information, then that's what the artist will make. This probably isn't that common though.

Generating textures in Photoshop after unwrapping the model is, to the best of my knowledge, still very common (though Photoshop might not be the particular picture-editor used). I've heard of an artist who uses GIMP to generate his textures.

As for terrain, the engine used probably supports its generation. For example, Unity 3D has a built-in terrain editor. All that must be supplied are the textures which will be painted onto the terrain. Another example is the engine that Saints Row: The Third was built on. A demo video showed a developer placing roads in the city, and the road segment being placed would automatically change to fit the road around it. Tools like this are implemented by the engine and tools programmers in order to streamline the development process. However, as someone pointed out in a comment, constructing these roads is really the job of the level designer.

Summary: The art pipeline is heavily dependent on in-house trade secrets, and the game engine. One of the only ways to learn the details is to enter the field.

Oh, and lastly, Pixar is well known for making public their advances and techniques in graphics, so they might have some information on their pipeline. It will vary from a game though, since they have the option of offline rendering.


There are articles around the internet. Some development teams post development blogs, it depends on who they're marketing the game to, it's often good marketing to show some "behind the scenes" magic.


CCP's Eve Online has one of the most geeky fanbase, they post alot of blogs on the technology behind running an MMO (beside the normal balance changes, they post on server maintnence, graphics, development team management, etc.). Their blogs are a great resource.



These are just examples, they probably don't relate to your question directly. I'd love to provide direct help, but i've never been seriously researching 3D technology, so i don't have anything specific bookmarked. I'm just here to tell you, the information is out there, you just have to dig it out.


You used to start off in your modeling package, and...

  • build base mesh
  • export to zbrush
  • sculpt and gradually add detail
  • export mid level detail back to 3d package
  • make low res model and cage
  • unwrap low res model
  • back high model to low using cage
  • test results in game then repeat last two steps until it looks right
  • texture

Or something like that. Every place is different, btw. Most studios have tools to rearrange or completely cut out numerous steps, because the process can be so crazy just for a single model of moderate complexity.


Not the answer you're looking for? Browse other questions tagged or ask your own question.