workflow problems
polycount lvl 666
So I have been kind of researching about workflow issues in zbrush/scupting in general and the main sticking point for me is that
say you 'finish' a sculpt and are relatively pleased with it, that's one thing - stage 1 achieved, lots of fanfare etc
but then a week later you start to think hmm it's not as good as I thought it was, so maybe you have to modify it quite a bit
(it could even be that a client might change their minds and wants you to totally rework a sculpt)
BUIT, by this time you may have baked out every map known to man, like ao , curvature etc, composited that in pshop and are feeling very pleased with yourself.
At the very least you may have to make slight modifications to your sculpt, but would still have to rebake all the maps and partially or fully retopo
your low poly model.
I think this is the main reason i don't use this workflow for most of my freelance work, ie I model stuff in 3d packages then texture it
which kind of limits the style of work I do.
I get the same money for doing both styles of work , but 1 of them takes 10 times longer. it's a no brainer really
I really wish that in the future this workflow disappears because its very time consuming and frustrating at times
If auto retopo could produce game ready meshes with good flow etc, that might be cool and huge time saver but I think its not really
there yet
any opinions or suggestions on this?
say you 'finish' a sculpt and are relatively pleased with it, that's one thing - stage 1 achieved, lots of fanfare etc
but then a week later you start to think hmm it's not as good as I thought it was, so maybe you have to modify it quite a bit
(it could even be that a client might change their minds and wants you to totally rework a sculpt)
BUIT, by this time you may have baked out every map known to man, like ao , curvature etc, composited that in pshop and are feeling very pleased with yourself.
At the very least you may have to make slight modifications to your sculpt, but would still have to rebake all the maps and partially or fully retopo
your low poly model.
I think this is the main reason i don't use this workflow for most of my freelance work, ie I model stuff in 3d packages then texture it
which kind of limits the style of work I do.
I get the same money for doing both styles of work , but 1 of them takes 10 times longer. it's a no brainer really
I really wish that in the future this workflow disappears because its very time consuming and frustrating at times
If auto retopo could produce game ready meshes with good flow etc, that might be cool and huge time saver but I think its not really
there yet
any opinions or suggestions on this?
Replies
But then you state that the time lost is in updating the lowpoly, bakes, etc. This is something you need to do whether you're creating the high with sculpting or with sub-d tools so... You've lost me.
well it reads fairly clear I think . I have been steering clear of the the whole, sculpt/retopo workflow in my freelance work because revisions take so long .
most of my freelance work is straight subd modelling with basic texturing, because its quicker, but the work I do is pre rendered on the whole, so not everything is baked in to the texture,( maybe i add a little ao) or a kind of fake lightmap. they are fairly quick to redo anyway
the point is the mesh I model initially is the final mesh, don't have to go through the whole retopo process
i do a lot of sculpting in my spare time , but loath to commit to it professionally as the process is so time consuming, major revisions
particularly - hope that's a bit clearer
maybe the base mesh is dense enough when the volume does not change significantly in the next subdivision up.
the problem is though, particularly with jeans and shirts is that the flow of the folds can be quite tricky to get right if you leave everything as loops/evenly spaced quads.
there would have to be a fair bit of retopo to catch the changes in direction, maybe even an intermediate stage of zremesher once you have the main shapes blocked in.
In that case you would definitely spend more time on your final low poly mesh. maybe in a big studio they would give you a proper timeline for these various stages, but I rarely get that TBH
It's odd though I have had some clients have n't even bothered with the retopo, just made me bake in to whatever the first subd level is.
can look a bit shit if the first subd is a lot different from the top level. not all clients have that polycount logic of doing it the right way.
yeah my main gripe was re revisions ie any changes you need to do to the sculpt involves redoing everything again further down the pipeline.
maybe its best not to do any ao baking until totally happy with the rest of it or even send the client a decimated mesh as a WIP - probably they would n't go for that
Sculpt in Zbrush. Use decimation or, more often, ZRemesher to get a 'quick and dirty' version of the asset into the game. If it really needs UVs, I use UV master. I don't bother saving the lower polycount version in Zbrush, just Maya, as I view it as temporary.
Once it's in the engine, I can easily revisit it in Zbrush and spit out a new temp/whitebox version in a couple of minutes(+ the amount of time to do the scultping). It can stay in this form for while, so I can think about the asset for days or weeks.
Once I know I really like the Zbrush version, then I do the 'permanent' topology and do the textures in Substance Painter.
I don't really do any hard surface or props, mainly characters BTW
When I did contract work in the past, I would always insist on getting approval in phases, for instance:
And if the client approved the UVs/bakes and wanted me to go back to the high, I would do it, but it would be very clear in the contract that they would pay for the revisions. This gives the client clear motivation to make sure the work is approved when it should be, and discourages trivial changes. Not to mention drastically cuts down on rework and wasted time for the artist.
This is all a bit easier if you're not doing any baking, but as you said, this will limit the sort of work you can do.
I seem to have a hit a brick wall here though and its been enough to put me off doing games work or work that involves the zbrush/retopo system.
rarely though does my subd work excite anyone these days, so its a bit tricky how to move forward
I think if I was in a studio that would be fine but as a freelancer they just want a certain batch of work for a certain price and don't much care for the process
I doubt this will be an entirely new concept to your clients either, a process like this is rather common. Unless you mean to say the majority of your clients give you a task and expect you to submit a completely finished asset when it is done with no communication during the task? Because that is rather uncommon in my experience, there is generally some back and forth, sharing of WIP images etc. What I'm suggesting is to make that more of a formal process so that progress and expectations are clearly communicated.
also the process of creating characters is not that clinical, I always ending up tweaking stuff throughout the process.
My clients are a varied bunch, they are not always like games people who have a very clear idea of the whole process re the games production methods.( particularly for hard surface modelling)
There is more of a loose arrangement re the process and even if you try and formalize the workflow, it doesn't always work out that way.
I am fine with my client process one the whole, it's the length of time/cost of actually doing zbrush work that bothers me, because the pipeline is so convoluted and revisions so time consuming .
Recently I had a character with sleeves down, so just the hands were poking out, but then they wanted me to change it so the sleeves were rolled up with bare arms.
Just modelling this, I did it in the one evening, but if that was part of a sculpting workflow it would have taken a lot longer with retopo/bakes etc. That is really my main argument here, that as a commercial artist, it's just too much effort/time to do the whole sculpting workflow and have a normal life.
shame because I love some of the top zbrush work, like the stuff from last of us etc
It's down to both parties not to be a jerk about it, ie a bit of flexibility goes a long way
You can't really start laying down the law too much or they will go elsewhere
It's a bit of a dirty little secret - artists sometimes like to claim that baking is a quick and easy step ("no problem, it will only take one day"), which it certainly is when everything is approved in a linear order (as it should be when everything goes exactly as planned), but isn't quite compatible with a more freeform approach to art and design.
It all sounds obvious to us, but I can totally see how a client might not fully grasp it. Especially since over the years the game asset creation process when from being totally open and non linear (ability to edit models and textures at any time, ability to perform UV tweaks on the fly, possibility for any artist or AD to edit textures directly) to being completely locked and linear (every step needs art approval, and every iteration past the baking stage requires the artist to step back, edit the model and UVs, rebake/retexture). From our perspective it comes with the territory ; but for someone not directly involved with the process, it understandably sounds like a serious step back compared to the much more forgiving workflow from a few years back.
A few solutions :
• Being super clear about the process, and establishing strict review points (but as said that depends on the client and its willingness to be involved every step of the way), and/or
• Only taking on jobs that do not require any sculpting/baking, and/or
• Hiring a contractor to take care of all the lowpoly/baking tasks, and/or
• Handling a model and its UVs in such a way that each part is strictly mapped to a distinct UV quadrant (square or rectangle) - therefore, if the sleeves need to be rolled up, the flesh and cloth parts can be edited individually (either cleanly rebaked, or hacked) without affecting the rest. Like here :
At the end of the day there's really no easy solution and I kindof wish artists were more open about these issues like you did here. As a matter of fact an artist able to develop an art style that looks modern while still allowing for non-linear tweaks is very likely to catch the eye of a knowledgeable art director interested in being able to tweak things on the fly. Arc System Works did pretty much this, and their work on Guilty Gear Xrd and DragonBallFighterZ is currently leaving Capcom in the dust since it looks a lot better than the bread and butter sculpted/normalmapped look of the upcoming MvC Infinite.
(The irony is that while CG artists are all about doing perfect sculpts in Zbrush, the public sees the end product and can spot the difference between a game looking cartoony but very clean and sharp and art directed to perfection, and a "sculpted" game where characters look like dead plastic dolls...)
I feel like the non-linear aspect of old school asset creation was a big part of what made us interested in it in the first place (it certainly was for me) so it's a bit of a shame that this facet of the craft is getting lost. There are ways to bring it back, but they do require to be proactive about it and do not apply to every job and art style (as opposed to being a core element of game art creation just a few years back).
my semi solution is to do partial re topos on my characters. so anything too difficult to model,( like certain kinds of cloth folds) I will quickly sculpt in the shapes and then retopo that section only .
So it will still be a subd model but the process of creating it would be a little more organized. still not a solution for real time stuff though
my best solution here is to bake 'only' the normal map in tests and keep the retopo very loose, because I like an approach where I can easily revisit stuff like cloth folds which never seems to be finished.
maybe need to get a really powerful pc so I can do the bakes faster
it still pains me though that the sculpt I am working on is not going to be the final piece, so the retopo aspect will always be a pain
ie build an evenly spaced quad mesh to sculpt on , sculpt it , zremesh it to get a nice flow , then build another version for your final mesh. what a real pain that is
at the very best and some clients want this - i build the final mesh with uv's then sculpt that in to shape - use the lowest level as the in game mesh and tweak it a bit so it matches the HP
the people you work for would really need to blindly trust your ability to deliver a satisfying result when left to your own devices.
i think it's quite common to tweak multiple aspects of the entire asset long after baking and sculpting. e.g. expression of the face directly on the lowpoly - effectively rendering the sculpt outdated. hacks, everywhere!
the only tip i have to keep workload down is to work from a uv-mapped basemesh that can double for game use. for humans this works well in my experience, just requires that time investment early on. then you can apply changes either on the sculpt or the game mesh and update the other with ease.
otherwise try to stick with medium res placeholder exports from Z for as long as you can and get approval for the entire character before really going in to optimize.
One of the things that a phase-based system does is force everyone in the process to fully flesh out their ideas. No half baked concepts or major details represented with a few brush strokes here.
"the people you work for would really need to blindly trust your ability to deliver a satisfying result when left to your own devices." This is why you hire talented artists. If you can't trust them to do quality work... Just, what? Thats exactly what they are being paid for. Additionally, with a proper phase-based approval process there is absolutely no "blind", the opposite in fact.
I understand that if you're actively designing the character or still nailing down the art style, you may want or need additional freedom, but this accounts for a small percentage of actual art production and is generally done with a less strict workflow anyway (blockouts, lookdev, unoptimized meshes checked in game, etc). Why character artists put up with this myth that the whole process needs to be fluid I have no idea. If someone changed half the design for an asset (a gun, a vehicle, a character - it does not matter) after I had the bakes done, I wouldn't say "oh my workflow isn't very good", I would call attention to the ridiculous indecisiveness of the process. Improving your workflow is always a good idea, but it's sort of missing the point in this case.
Character art isn't some magical, etherial thing that can't be defined with concept art and reference. I've heard countless horror stories about characters being redesigned for trivial reasons after being fully rigged and animated, costing studios tens of thousands of dollars per character. With a sane approval process, this problem would go away almost entirely. This isn't a workflow problem, it's a management/decision problem. It's true that the individual artist may not have much say in this process, but to insist on it or look at it as the ideal way to work simply perpetuates the myth and causes more problems for everyone involved.
Major changes in production assets should be the exception, not the norm. If this is happening with every asset you do, you have bigger problems than how long it takes to rebake.
I'd totally agree that ideally there should be no retakes and things should progress linearly ; but as @thomasp said, making a character "gel" often requires tweaking a fair amount of subtle things on the fly, even when the provided concept art is as tight as possible.
Personally the way I tackle this issue for characters (when working on a model myself, or when working with a dedicated modeler) relies on making the process as wysiwyg as possible. This includes :
- Never, EVER reviewing anything based on Zbrush screenshots. The perspective is off, and the viewport controls don't allow for smooth inspection anyways. Screenshots should be taken in Max/Maya/Blender, or in-engine, or in TB3. I also tend to be wary of self proclaimed "Zbrush wizards" ("This mech was made 100% in Zbrush !"), because that means they spend a majority of their time in an environment not representative at all of the final destination of the asset (and therefore the perception they have of their own work is off at all times).
- Always asking for models to have simple colored materials applied to their various chunks. This forces the artist to split everything in parts, thus making sure that everything will be ready for an ID bake. And of course it provides visual parity with the concept.
- For sculpted faces, always asking to have a light polypaint/texturing pass done on the high, to match the intended goal as closely as possible. Because this http://www.hauntedstudios.com/prodimages/joliebustwhiteG_medium.jpg
is not as useful as this
https://s-media-cache-ak0.pinimg.com/736x/8f/3b/9c/8f3b9cda3783cf34a8354d7122e4a91c--james-white-jolie-pitt.jpg
I am personally comfortable with this review process but it's clear that is is nowhere near as fluid as what was possible before baking. As said, it comes with the territory but I personally cannot blame Art Directors for not knowing the intricate technical reasons why things can now slow down to a crawl when asking for a seemingly small change at the end of the asset creation process (like rolling up the sleeves of a character) - compared to the full editing freedom that was possible a little while ago.
Now on the topic of management : I do agree that there are issues, but some of them come from the artists themselves. For instance, many artists insist on jumping straight to sculpting, because it's the sexy part that looks cool in screenshots and being able to show a cool sculpt in just a day or two seems impressive. But that means that the rigger (or client) will have to wait a few extra days to get a good proxy model, and so on. Now of course the people in charge of management should know that and ask for things to be done in a clever order (blockout/proxy first, so that it can be passed to animators while the polished sculpt is being done) but the issue is that it now has to be a proactive decision that artists sometimes resist, whereas before it came naturally.
either that or when you work for people who don't really put a lot of emphasis on characters (which probably means the publisher is entirely hands off) or are so budget constrained they'll have no choice but agree to everything.
in any other case however with characters being a focal point you'll have the situation where everybody and their dog has an opinion on them and some of those opinions will carry enough weight to trigger rework. be it a likeness they want changed, a case of "not sexy enough", not liking some aspect of the concept but they had to start the model anyway to make a deadline or whatever.
also that'll be a reasonably complex asset with knock on effects for body deformation, physics, facial animation and whatnot. you can mock things up to some extent to get those responsible a test case but ultimately and especially for facial you'll likely be back tweaking topology or replacing entire chunks at some point.
and i don't think it's the norm to rip characters apart late in the game but rework here and there including replacing a head - definitely. i've not been on any project where in the end the assets weren't pulled together to be judged side by side and lots of tweaking ensued.
- Blockout for major proportions
- Periodic updates to check how well forms read in HP model - most major design choices are nailed down here
- Basic material colors and shader values plugged in - previewing with a skin shader gives you a great idea of how your forms will hold up for character work, adjusting lighting gives you a good idea of how the asset will look in different environments - further design tweaks are easy to do here - if you've got an indecisive AD, don't move on until they are happy with this stage
- Lowpoly, check proportions and silhouette
- Messy uvs, test bake to make sure bake works well/details read correctly
- Final bakes with properly packed uvs
- Materials
I make countless tweaks with this process, it's really easy to spot how well forms read, if your detail scale is off, etc. By the time I have my highpoly done, I essentially know exactly what the lowpoly + base materials will look like. Checking in game is an option too, though generally more work involved (decimate first?).Now specifically to a linear, phase based approach, the idea that you can't make tweaks to fine tune the design is a misnomer. You simply want to make tweaks at the right stage. Again having a representative preview of your highpoly model helps LOADS here. You can make the vast majority of your design-important changes here. If you need to move a few verts around after the final bake that's generally easy to do (even if you need to tweak both the high and low and rebake).
Basically, you don't commit to the lowpoly stage until you're sure the design is what you want, and again this is not a terrible crime or impossible to do, it actually simplifies everything quite a bit when you start doing it.
As far as ADs not really understanding the technical drawbacks of asking for changes late in the process, sure, I get that. Which is why it's all the more important to talk about it here, and have conversations at work!
Firstly, you have the sort of situation you describe above, typical for central characters of a game. Someone high up wants a new face or a new outfit or a drastic redesign of some sort. This is inevitable and will tend to require significant rework no matter how you approach it.
Secondly you have more trivial things that will pop up in all manners of production (from the smallest prop to main characters) if you don't have a solid approval pipeline. That's mostly what I'm talking about here, and much of this can be easily avoided.
Pretty much like what Joe wrote, with possible extra steps inbetween. Some clients wanna see the UVs, some want to see the asset baked before they judge this step. Things like this may vary, some productions might start with the low which then gets subdivided and detailed, others will be more classical, with just handpainted textures on lowpolies.
But in any case, you should make sure, your clients know your deliverables and approval steps, which have to be strict to a certain point. Otherwise the both of you will have to run through crazy amounts of interation.
well, Neox ,seemed a bit like you were having a bit a of a 'pop' at me
my clients are not all game guys. some are advertising, some are toy companies. i have worked ion jobs where the have basically told me to spend 5 weeks working out good techniques . One size does not fit all
I totally understand the standard approval process and am trying to say here that not every adheres to it.
hell, some clients of mine are not even 3d guys really, so you can't be a total 'twonk' when dealing with them.
most of my recent games work had a fairly loose arrangements in terms of schedule and fixes. generally worked ok for us, but was really just
]ust asking for advice for the zbrush/baking aspect in this thread. ie the fact that the sculpt workflow is quite time consuming and fixes even more so.
it's a really big thing for me because it totally affects the kind of work I do.
I rarely use photoshop anymore.