Achieving realism: Handpainting Vs. Photosourcing
polycounter lvl 19
This is derived from the Level Design Vs. Env. Artist thread. The topic here is creating textures for realistic art themes.
Here's a quote from one of the replies from the above thread to get this discussion going.
In my opinion using CGTextures or other libraries is a colossal time saver. Lots of companies - the majority of them - rely on quick turn around for quality art. Highpoly modeling, baking, then "hand painting" your textures is a colossal time sink. This simply is not a fast process, no matter how good you are "photo sourcing" will always be the fastest result. Is it the best result? Maybe it is. Maybe it isn't. It depends on the schedule your game has and what the art direction calls for.
When I say "photo sourcing" I do not mean simply copying/pasting a texture from CGTexture, Merlin, and so forth and calling it done. When I say photo sourcing, assume that I am talking about using a photo as the base for your texture and going upward and outward from there to achieve the texture you need. Yes I understand that some games simply c/p their textures but even then thats not how the majority of their textures were created.
To me, sculpting out every texture isn't necessary to achieve 'quality' art. If you use a texture library over and over again where you're making changes here and there as you go it won't mean shit to the games final result. No gamer is ever going to pick up on that and really the only thing being challenged by it is your ego and pressure from your artist peers.
Now to completely contradict myself: I wish I could get paid to sculpt every realistic texture, bake it down, bake out the AO, then hand paint. The end results would be more rich, technically accurate and 100% owned by that game. However, I've shipped 7 games now and not once did the schedule or budget allow for such an approach to texture creation.
My point? Both processes have their place but the reality is it's different from project to project, schedule to schedule and budget to budget. Educate yourself on both techniques but just be sure that either process is something you're knowledgeable about, efficient with and are able to create quality artwork from.
This is strictly my opinion and not the word of the Game Art Gods. I'd like to hear everyone's take on this - especially those of us busting out artwork professionally. Every studio has their process, whats yours?
Here's a quote from one of the replies from the above thread to get this discussion going.
Lord, if I see one more time someone using CGTextures and getting away with it that's not a student/college based first semester model, I'm going to punch a polar bear.
In my opinion using CGTextures or other libraries is a colossal time saver. Lots of companies - the majority of them - rely on quick turn around for quality art. Highpoly modeling, baking, then "hand painting" your textures is a colossal time sink. This simply is not a fast process, no matter how good you are "photo sourcing" will always be the fastest result. Is it the best result? Maybe it is. Maybe it isn't. It depends on the schedule your game has and what the art direction calls for.
When I say "photo sourcing" I do not mean simply copying/pasting a texture from CGTexture, Merlin, and so forth and calling it done. When I say photo sourcing, assume that I am talking about using a photo as the base for your texture and going upward and outward from there to achieve the texture you need. Yes I understand that some games simply c/p their textures but even then thats not how the majority of their textures were created.
To me, sculpting out every texture isn't necessary to achieve 'quality' art. If you use a texture library over and over again where you're making changes here and there as you go it won't mean shit to the games final result. No gamer is ever going to pick up on that and really the only thing being challenged by it is your ego and pressure from your artist peers.
Now to completely contradict myself: I wish I could get paid to sculpt every realistic texture, bake it down, bake out the AO, then hand paint. The end results would be more rich, technically accurate and 100% owned by that game. However, I've shipped 7 games now and not once did the schedule or budget allow for such an approach to texture creation.
My point? Both processes have their place but the reality is it's different from project to project, schedule to schedule and budget to budget. Educate yourself on both techniques but just be sure that either process is something you're knowledgeable about, efficient with and are able to create quality artwork from.
This is strictly my opinion and not the word of the Game Art Gods. I'd like to hear everyone's take on this - especially those of us busting out artwork professionally. Every studio has their process, whats yours?
Replies
A truly good game artist can use CGtextures without anyone being able to tell. Simply because they've done a lot of artistic choices to progress it past a simple 'copied texture'.
-2 more cents
You can get into some pretty big legal issues if you use images you pilfer from any old website so you need to make sure you're not putting your company at risk. Which is why sites like CGTextures are so popular.
all people is using the same source for photo-textures? xDDD
That it was. Beautiful and fun game too!
For those who dont want to read it is mainly linked to Art Direction above all else. Uncharted is not a photo realistic game, it is stylized realism.
Chances are your audience doesn't care and artists are not your audience. Your audience is probably not even going to notice; how many players stand around looking at every texture in detail? Are you getting paid to please your peers in the art community or make a game?
Plus, consistency in art direction trumps everything else. Look at HL2! Photosourced to hell and back with lots of sloppy assets but still looks great as a whole because the direction was consistent.
Its your job as an ARTIST to know how to put those visual languages together to capture a particular experience that either you or a concept artist has delinted.
You need to know what is required to evoke the desired effect in the player then take the shortest possible route to see the result in game. Then you can start the iteration and refinement process that takes into account the many constraints an environment artist faces.
Its about KNOWING what to do, not HOW to do it.
Knowing how to do something is the job of a draftsmen. This is someone who just copies and reproduces things with great efficiency and speed in a particular way with consistent results.
Environment artists have to be both skilled artists and skilled draftsmen, but you should never put your role as a draftsmen before your role as an artist.
I have insane respect for those who can completely hand paint convincing textures, but sometimes it's just not reasonable, and also, I just don't have the painting skills necessary to do that.
I love this thread.
2) You grab 5 photos and cut these up to create one compleet new texture.
3) Use brushes only to create your texture.
4) Mix techniques
Non of these catagories is wrong. They are all applied in good time. May it be because of a deadline, or because it gives the best result, the fastest, the fitting style, so on.
Now see that they can also be mixed with,
1) Something that is made at work
2) A learning experience
3) A display of skill
Yeaah.. so wait where were you pointing your finger at?
You know I love details, and I have quite some experience chasing that by brushes and smudgetool only, trying to understand and getting those photorealistic details in. To just finish it with a nice noise pattern or create a dirt layer to scatter the surface. My bad, did i just sample something from another source.. even if it is generated..
It is a valueable skill to be able to paint propperly, cause you cannot hit all shades you need in a texture by just cutting up photos only.
Though you can also not hit the surface details without sampeling something.
But somewhere down the line it will mix.
And if you have experience and understanding of how details work and how you can go around to create them, you will find the fastest AND best looking result.
and what Jacque and other mentioned
Thanks for mentioning that, it lead me to search and I found this: http://www.cgtextures.com/content.php?action=tutorial&name=blendif
Process is completely meaningless. Do what works best.
I know guys that can hand paint the most realistic textures imaginable in the time it takes to manipulate a photo. They have nearly a decade of gathering and creating photosourced brushes, and have a library that contains every kind of brush imaginable to create any kind of texture. They simply have that foresight and experience.
Who knows, maybe one day we'll all be using procedurally generated textures. Who cares how its made. You're gonna have to relearn aspects of your trade at one point or another.
Quality + Efficiency.
But surely there were photos involved when creating new brushes for photoshop or brush-patterns for zbrush?
Not to say naughty dog didn't go full on handpainted, since I still would consider it exactly that, I just think that photosourcing is involved more or less everywhere.
And as people have said earlier in this thread: as long as the end result is great.
It does feel like the programming-worlds discussion on which language is the most proper one to use, and how it affects the end-result of the game.
2cents
You'd be a fool not to use photo reference at times, even to just to give you a good blueprint to start from.
I start with AO, material mask, and normal map bakes.
I begin creating layer sets for each material present in the texture.
I then fill this with a base color, painted texture (I have a sweet brush that paints instant speckled rubber, for example), or photo, and tweak/crush colors and value until I have a solid start. I then mask this layer set off and paint it into where it belongs in the texture (with the help of my baked masks). Sometimes this layer set may contain several layers of painted corrections, color adjustments, painted or photo-pasted details, etc.
I then nest a second layer set for the sub-material - for example, if I have painted metal, I will start with a painted metal base, and then make a subset for scratched metal. I'll fill this with a clean metal base photo, color, or painted texture. I'll then mask the whole thing off, and begin painting in the scratches and nicks into the clean_metal layer mask.
I'll continue to nest layers down as needed - for example, scratching through the paint might reveal fresh metal, or it might reveal old metal, which is a sub-material of the fresh metal. From there, the old metal might have new rust, old rust, etc.
Going back up, the paint might have sun fading affecting saturation, etc.
I'll repeat this for each material until I've filled out my diffuse. Then I group it together, copy/paste it, and rename it specular - this preserves my layer set hierarchy and masks. From there I'll alter/replace/embellish each layer set's worth of materials.
I'll do this again for reflection, gloss, etc maps.
The best thing about this system is that once I'm completely done, I've now got a great template texture that I can use for related textures - all I have to do is blow away the layer masks and paint new ones and I'm 90% of the way there.
In the end, I wind up with a texture that relies on handpainting for the details and 'finish' while using photos and hand painted elements as needed, and serves as a useful tool for future textures to ensure consistency among related textures.
If I was doing a lot of the more organic environment stuff the method Autocon talks about seems really good for generating strong normal maps that correspond with the diffuse/specular. With the hard surface stuff the painted details that have corresponding normal map information are usually subtle surface treatments, etc, that can be filtered off a photo or painted heightmap easily, which isn't always the case for rocky hills, cobblestones, dirt, etc.
Very true, photos are used for brushes/brush patterns im sure.
I shouldnt say 100% of everything was totally painted in every aspect as nothing is ever done 100%, 100% of the time.
As many have said its not which is better, the end result for your time frame and ascetic look is what is important.
The stylized realism look is look for Uncharted. It also helps for the creation of better normal maps, creation of maps we need for blending, easy when we need to make changes to textures and the ability to share ztools for faster texture creation.
Where as at Bungie I used photosourcing for Halo Reach stuff as doing things the way Naughty Dog dose for Uncharted wouldnt work for Halo. It all depends on the game.
At Insomniac is was really up the artist. They gave us a lot of freedom. The only thing that was important was that the end product was bad ass. I used my own account at CGTextures and my own portable HDD of textures I've collected over the years. But we did have a folder somewhere..deep in the bowels of our servers where there were a couple of textures we could use if we wanted.. ie: Blood Splatter, Wood etc. But I never had any use to use them. . .