I dabble in blender. I've noticed as time has gone on its going to be hard to filter out outdated or impractical methods from newer more efficient ones. It's not really anyone's fault, some shortcuts that work well on a simple flow will mess things in the long run for bigger projects.
Aside from the comments sections, which will crucify a bad tutorial, how is the laymen to know the difference?
This is a problem with tons of applications. I think that's another strike against video tuts: Yes there are outdated text documents, but it's far easier and more likely that written instructions will be updated/amended than it is for a video.
It's always more of a problem with "accessible" software like Blender. Next to no barriers to entry results in too many content creators and too little knowledge. Couple that with the fact that Blender hasn't yet become an industry standard, and there just isn't the same kind of tuition available.
Meanwhile convoluted and expensive industry staples like Houdini have almost consistently good content. I'd wager for a lot of the more theoretical topics, which are largely software agnostic, you'd be better off picking up the basics anywhere and then moving onto something more reputable for the theory and advanced topics even if the course uses another software package.
This overlaps somewhat with my interest in modelling for 3d printing, and I suspect that in some cases we'll see better pure modelling coming out of that community than the gamedev side - if nothing else, a badly clipped edge can be hidden in the game, but it turns into unusable garbage when you try and slice it for printing.
Saying that, I've spent more than a year now on one model and not even tried printing it yet, so I am not your guy for efficient workflow.
but I feel like that's for people who are absolutely anal about polygon count
Nah, automated tools are great at hitting polycounts. They're just terrible at providing adequate density and correct edge flows in light of intended deformation leading to poor results on animated/deforming meshes. Looking at part 4 where the retopo starts, that looks like fairly straightforward manual retopo. Admittedly, I skipped through it owing to its long playtime and relatively low quality.
In terms of tools, I use a couple to speed up the process. ZRemesher with guide curves can get pretty close for unimportant meshes where turnaround time is paramount. For hero characters, where I want complete control over flow and density, Topogun 3 makes it less of a chore.
For unconventional characters (creatures, weird proportions), I work freehand without thought of topology in Zbrush. Once the form is pretty close to final, I'll retopo and unwrap before projecting and finalising the sculpt. In the end, your lowest sub-d level is essentially your LP. You'll often want to make some minor adjustments before bake, but they are that, minor.
Generally though, you should be looking to avoid as much topology and UV work as is humanly possible. It's a time sink, a formality required to have nice, consistent characters.
Reshaping the closest candidate from a set of base-meshes I've made, and using projection workflows is something I do wherever I can. Best case scenario? Tweaking some verts in a couple of tricky places and relaxing a couple of interior UVs after sculpting. Minutes. It also makes for relatively consistent UVs, which you can use to accelerate skinning across similar meshes.
Polycount is a small part of the motivation for good topology in regards to characters.
The flow of edges determines how the surface will change and shade in response to deformation. Bad topology can absolutely ruin the best looking character when animated, while making skinning/rigging a pain in the ass and unnecessarily time consuming. If you're working in a team environment, this becomes more important. Any rework incurs latency and wastes team time, and you'll invariably piss off your colleagues.
That said, your artistic requirements will determine how important this is - you're unlikely to notice issues on a RTS units face. A realistic character viewed up close in first person? Definitely a problem. Basically, as little concern as your project allows for, but no less.
That in mind, since I tend to do more detailed characters, the best approach I've found is the one that allows me to avoid any real retopology or unwrapping for the majority of my characters (clothing excluded, but a lot of that can be automated) giving me the time to either push quality elsewhere in the pipeline, or spend it on characters that don't allow for such an approach.
Also, extrusion is really slow. Blender got PolyBuild not too long ago. It's similar to Maya's QuadDraw, both of which are inferior to Topogun. Seriously, check it out. Really quick.
Right? Take a look at ZRemesher as well - it's still the fastest consistent workflow I'm aware of, though there are cases where you'll want to do it by hand, and that's where Topogun comes in. You'll want to do some clean-up around loop intersections for animated meshes, but considering game models can exceed 100k polygons these days, getting 98% of the way there in a couple of minutes by drawing guides is pretty great.
I really like sculpting and I feel that's ideal for making detailed characters compared to block modelling
Absolutely. ZBrush was revolutionary when it came out, and a large part of why CG took a huge leap forward in the early 2000's. Polymodelling has it's places, but organic detail is not one of them. Doesn't matter whether it's detailing chipped concrete or warped metal, producing gnarled trees or characters, if you want your art to look good outside of "lowpoly" art styles, I'd say a decent grasp of sculpting is absolutely mandatory. It relies more on traditional art skills than conventional modelling, but as you get better you'll find yourself able to sculpt complex characters without references in a couple of hours. Really rewarding being able to make anything you can think of, quickly.
I'd honestly suggest trying Zbrush. It's unfortunately moved to a subscription model, which I'll have to decide on when it gets an update I actually care about (my permanent license no longer receives updates), but nothing comes close in terms of sculpting. It's also pretty great for hard surface modelling where the resulting topology isn't especially important (almost all of it).
Expected thread about high performance asynchronous data buses in highly parallelized game applications.
Got thread about how stupid artists organize their stupid work.
Immensely disappointed.
Re: the question of the unreliable tutorial
I dabble in blender. I've noticed as time has gone on its going to be hard to filter out outdated or impractical methods from newer more efficient ones. It's not really anyone's fault, some shortcuts that work well on a simple flow will mess things in the long run for bigger projects.
Aside from the comments sections, which will crucify a bad tutorial, how is the laymen to know the difference?
This is a problem with tons of applications. I think that's another strike against video tuts: Yes there are outdated text documents, but it's far easier and more likely that written instructions will be updated/amended than it is for a video.
It's always more of a problem with "accessible" software like Blender. Next to no barriers to entry results in too many content creators and too little knowledge. Couple that with the fact that Blender hasn't yet become an industry standard, and there just isn't the same kind of tuition available.
Meanwhile convoluted and expensive industry staples like Houdini have almost consistently good content. I'd wager for a lot of the more theoretical topics, which are largely software agnostic, you'd be better off picking up the basics anywhere and then moving onto something more reputable for the theory and advanced topics even if the course uses another software package.
This overlaps somewhat with my interest in modelling for 3d printing, and I suspect that in some cases we'll see better pure modelling coming out of that community than the gamedev side - if nothing else, a badly clipped edge can be hidden in the game, but it turns into unusable garbage when you try and slice it for printing.
Saying that, I've spent more than a year now on one model and not even tried printing it yet, so I am not your guy for efficient workflow.
You too huh?
Tbf I've modelled and printed a bunch of stuff in the meantime, but there's this one project that's taken way too long, lol.
Nah, automated tools are great at hitting polycounts. They're just terrible at providing adequate density and correct edge flows in light of intended deformation leading to poor results on animated/deforming meshes. Looking at part 4 where the retopo starts, that looks like fairly straightforward manual retopo. Admittedly, I skipped through it owing to its long playtime and relatively low quality.
In terms of tools, I use a couple to speed up the process. ZRemesher with guide curves can get pretty close for unimportant meshes where turnaround time is paramount. For hero characters, where I want complete control over flow and density, Topogun 3 makes it less of a chore.
For unconventional characters (creatures, weird proportions), I work freehand without thought of topology in Zbrush. Once the form is pretty close to final, I'll retopo and unwrap before projecting and finalising the sculpt. In the end, your lowest sub-d level is essentially your LP. You'll often want to make some minor adjustments before bake, but they are that, minor.
Generally though, you should be looking to avoid as much topology and UV work as is humanly possible. It's a time sink, a formality required to have nice, consistent characters.
Reshaping the closest candidate from a set of base-meshes I've made, and using projection workflows is something I do wherever I can. Best case scenario? Tweaking some verts in a couple of tricky places and relaxing a couple of interior UVs after sculpting. Minutes. It also makes for relatively consistent UVs, which you can use to accelerate skinning across similar meshes.
Polycount is a small part of the motivation for good topology in regards to characters.
The flow of edges determines how the surface will change and shade in response to deformation. Bad topology can absolutely ruin the best looking character when animated, while making skinning/rigging a pain in the ass and unnecessarily time consuming. If you're working in a team environment, this becomes more important. Any rework incurs latency and wastes team time, and you'll invariably piss off your colleagues.
That said, your artistic requirements will determine how important this is - you're unlikely to notice issues on a RTS units face. A realistic character viewed up close in first person? Definitely a problem. Basically, as little concern as your project allows for, but no less.
That in mind, since I tend to do more detailed characters, the best approach I've found is the one that allows me to avoid any real retopology or unwrapping for the majority of my characters (clothing excluded, but a lot of that can be automated) giving me the time to either push quality elsewhere in the pipeline, or spend it on characters that don't allow for such an approach.
Also, extrusion is really slow. Blender got PolyBuild not too long ago. It's similar to Maya's QuadDraw, both of which are inferior to Topogun. Seriously, check it out. Really quick.
Right? Take a look at ZRemesher as well - it's still the fastest consistent workflow I'm aware of, though there are cases where you'll want to do it by hand, and that's where Topogun comes in. You'll want to do some clean-up around loop intersections for animated meshes, but considering game models can exceed 100k polygons these days, getting 98% of the way there in a couple of minutes by drawing guides is pretty great.
Absolutely. ZBrush was revolutionary when it came out, and a large part of why CG took a huge leap forward in the early 2000's. Polymodelling has it's places, but organic detail is not one of them. Doesn't matter whether it's detailing chipped concrete or warped metal, producing gnarled trees or characters, if you want your art to look good outside of "lowpoly" art styles, I'd say a decent grasp of sculpting is absolutely mandatory. It relies more on traditional art skills than conventional modelling, but as you get better you'll find yourself able to sculpt complex characters without references in a couple of hours. Really rewarding being able to make anything you can think of, quickly.
I'd honestly suggest trying Zbrush. It's unfortunately moved to a subscription model, which I'll have to decide on when it gets an update I actually care about (my permanent license no longer receives updates), but nothing comes close in terms of sculpting. It's also pretty great for hard surface modelling where the resulting topology isn't especially important (almost all of it).
I have really hated videos. I want text so then I can go at my own pace. Even AI tutorials have been more useful.
Looks useful, I check those out later. Thanks.