This is so crazy - the sky is larger than the sun - the sun is thousands and thousands of times brighter than the sky.
Yeah the sun is brighter, but the sky has just as much influence because of its size. Stand in the shade, anything that isn't touched by the sun or is shadowed, these areas are light by the sky.
The sun isn't always around is it? But the sky is.
ark: Let's not hog this thread. Use the search feature or google it. This subject has been done to death. Like I said: the specular map defines properties of the materials, not of the environment.
Ok, im sorry for derailing this great thread. I'll read up on what you say and have a rethink on what i have said.
Ok, understood. I don't think people doubt that natural ability exists. Everyone must understand that you can be born with fingers that are suitable for playing the piano or legs that are suitable for running fast. The real problem with the term "talent" is it's often used by people who have no idea what it means and assign supernatural attributes to it, or are too specific. Like there's no "drawing" talent or "scuba diving talent", but there's natural levels of motor control, visualization pathways, lung capacity and so on. Small contributions in any case. I consider it an insult to tell people they have talent. It's like saying "oh this was easy for you to learn, you were just born with talent" as if they haven't worked their asses off for what they can do. How do you even know that someone has "talent"? Maybe they had the opposite of talent. Maybe they had to fight their very nature to get good at something and it took them years of torturous struggle and then someone comes along.. "oh, you're talented".
And that right there ladies and gentlemen is a thing of beauty. Too bad that doesn't fit on a bumper sticker, a business card maybe.
Add to it the technical dimension that EQ talked about and you have a super complex skill set that didn't just fall out of the sky. I could almost go along that art is easier for some people (there is a long list of why that could be that doesn't relate to some mysterious art glad that is yet unidentified) but even then they still have to learn a mother truckload of technical skill to make their art functional. Which for most artists is counter to how their brain normally operates.
Someone calling an artist talented doesn't necessarily piss me off, it depends on the way they say it. When someone starts throwing around gift or gifted then that starts to grate on my nerves instantly.
Ok, understood. I don't think people doubt that natural ability exists. Everyone must understand that you can be born with fingers that are suitable for playing the piano or legs that are suitable for running fast. The real problem with the term "talent" is it's often used by people who have no idea what it means and assign supernatural attributes to it, or are too specific. Like there's no "drawing" talent or "scuba diving talent", but there's natural levels of motor control, visualization pathways, lung capacity and so on. Small contributions in any case. I consider it an insult to tell people they have talent. It's like saying "oh this was easy for you to learn, you were just born with talent" as if they haven't worked their asses off for what they can do. How do you even know that someone has "talent"? Maybe they had the opposite of talent. Maybe they had to fight their very nature to get good at something and it took them years of torturous struggle and then someone comes along.. "oh, you're talented".
This is interesting. I never heard about this point of view before, but you are right. If you look at it from this perspective then it's somewhat of an insult. I don't want to say that people call me talented very much, but if they do I feel grateful for the special and strange situation where I've grown up. It remembers me of beeing an individual and how I became what I am today.
Edit: This is getting slightly offtopic, sorry for that.
your example is good except for one thing - your model is made up of multiple objects that share a texture - if they are seperate objects you only need a greyscale map and a color swatch to modulate the spec and get the same effect. the option to use rgb spec textures is great from an artists perspective and I prefer it if its available.
I found that learning how to use Blender greatly improved the way I think about topology. Learning how to model a "different" way. Before I started Blender I was simply sticking to my student licensed 3DS Max, and a lot of times it felt like I was more of a 3DS Max user than an actual CG modeler. Blender was so different that now that I know it I find that I can switch to any 3D package with ease.
Plus Blender's shortcuts are hard to learn but invaluable when you do as it allows for some very quick modeling. I still primarily use 3DS Max because of it's status as industry standard, but Blender's usefulness as a speed modeler is not to be scoffed at.
-When the UV layout is done, they scale up all pieces non-uniformly to use the remaining space: This can make the pixel density inconsistent, which looks bad.
-They'll tweak the island shapes to stack better: And in the process introduce awful warping and stretching.
Eh... I dunno. I'd disagree there a bit.
I scale up bits and pieces if I've packed it in and there's a few little percent I can squeeze out of the remaining space. Or if some part of the model is so small and I know I want a certain amount of pixels on it for a particular detail. The difference in texel density shouldn't be noticeable if it's packed sufficiently well at 1:1 texel ratio beforehand and it gives you a bit more texture space to play with.
And I'll tweak island UVs to have straighter/aligned edges. Usually the stretching isn't noticeable and it's a lot easier to texture. Makes them easier to pack, too.
Mostly just pointing out that there's a considerable gap between "inconsistent" and "bad". Sounds a bit anal, I guess, but if this is a thread about clearing up myths it might help to mention it's not 1:1 texel ratio or nothing.
You can get away with skewing the density a fair bit. It's when it gets that some parts have twice the texels of other bits that you're taking the piss and your UVs suck.
I dont think Per was saying that your uvs need to be 1:1 consistant over the whole mesh, its when you start scaling things up to fill large spaces, instead of packing differently or not worrying *so* much about using ever pixel. When people get obsessive with using every pixel bad things tend to happen, texal inconsistency, not enough padding resulting in poor mipping, etc.
It is very important to keep in mind the math that goes into scaling uvs. For instance
125x125% = 1.5 pixels for every 1
150x150% = 2 pixels for every 1
200x200% = 4 pixels for every 1
So even scaling something 125% can make a rather large difference. I think people make the mistake of thinking that when you scale something 125%, its only 1.25:1, instead of 1.5:1, so this can easily get out of hand.
Scaling things +/- 10-15% or so can be fine, but when you get higher/lower than that you need to be very careful.
Also, if you're scaling up AND down to fit everything in, you get your consistency out of whack really fast. Scale one thing down 90%, scale another up 115%, and all of sudden one piece has pixels 1.5 times bigger than the other, not good.
Jedah, I dont remeber anyone saying "Learning Blender is a waste of time", or am I missing something?
I mean a dumb thing would be to learn blender only, then hope to be up and running day one at a professional studio not using it in its pipeline (for very specific and justified reasons other than fanboyism hehe). Of course if the portfolio is steller it wouldnt really matter to get the job, but in order to keep it one would likely have to learn another app, thats just a reality.
Regarding this - I would be very interested to see more Blender myths not necessarily debunked, but at least put to the test. For instance, its modeling toolset. Of course die hard Blender-only users swear by it ; but I would love to see a skilled Blender artist perform a series of simple and complex modeling tasks on video in order to tell if, indeed, Blender is an efficient modeling environment. In this sense, Blender CG movies don't prove much - they show that the end result can be great, but they say nothing about what it took to get there!
So you seem to be sold regarding Blender being a faster modeler than Max. I totally trust you in the sense that it seem to fit your workflow and that's what matters. Could you be a bit more specific? Maybe pointing to videos showing what you mean. If it feels like too much of a derail you are free to PM me.
(I actually do want to switch my modeling app to something else than Max but so far not a single one topped it, after extensive testing. I keep trying out Blender regularly but some things are just backwards for no reason, and I cannot trust a software showing the obvious signs of stubborn project management ...)
Yeah the sun is brighter, but the sky has just as much influence because of its size. Stand in the shade, anything that isn't touched by the sun or is shadowed, these areas are light by the sky.
The sun isn't always around is it? But the sky is.
Yes. Consequently, if you get a mirror and set it on the ground there's going to be a hell of a lot of soft blue and a super bright sun reflection.
But skin isn't near reflective enough to send any of that blue color torward the eye, only the very brightest lights are going to bounce off and make it all the way to us -- artificial lights near the surface, and the sun are bright enough to cause highlights on the skin.
And like per said, specular defines the object. Throw a grey spec on a grey object and fill the scene with bright green lights, you will get green highlights.
This has limited scope, blender while being extremely capable has a limited install base in studios compared to max or maya. If you have the choice id say learn a big name app, if your working in house you'll probably not be using blender so after getting used to blenders rather unique way of doing things you need to learn the other way too..which is a pain (I had to). You can get an edu license of maya of max for free at the moment so id push newbies in that direction myself. not because they are more capable but purely because they are more widely adopted in production.
The polycount is just one part of an asset that you want to keep under control to help efficiency and usability, but triangles are very inexpensive compared to the alternatives. There are many factors to be considered when optimizing a lowpoly mesh outside of simply affecting the silhouette. Will adding some extra verts allow you mirror portions of a texture and possibly drop the size of the texture? Vertices can also be evaluated for their aid in lighting, collision, or deformation if the model is to be animated. Overall though, often details on a model or bevels on an edge can help the model much more than watching your tri count drop by 20 ever could, especially when it comes to normal-map driven assets.
That's hardly a myth. On high end hardware the triangle budget is huge of course, but even so, there's a limit.
And on hardware like the DS where there's a hard limit of 2000 per frame, then the triangles are not inexpensive at all. It just depends on the platform.
This is pretty bad advice in almost every context, however it gets repeated probably more than any other single myth. Here are a few points that should cover most situations in which you should/shouldn't use triangles/ngons:
....
2. When it comes to highpoly sub-d work, it is true that it is best to avoid triangles or ngons in high detail areas where shapes are twisting and turning, on convex or concave surfaces. However, it is perfectly fine to use them on flat surfaces, or unimportant areas that are likely never to be noticed. Ngons can actually resolve better in some special cases than quads as well, so experimentation is key here. Again keeping all quads for the sake of doing it is just silly.
3. When it comes to creating cages for sub-division sculpting (zbrush, mud, etc) this is where this "Myth" actually has merit. Triangles can cause weird issues and bugs in sculpting apps, and sculpting detail on triangles can result in bad visual artifacts as well. When it comes to sculpting, its best to use sensibly flowing, even shaped quads, wherever possible.
4. Ngons can be give unpredictable triangulation, so using them in excess, and not cleaning them up before the model is finalized can give you some problems. However, being anal retentive about forcing quads while you're modeling is only going to slow you down.
While I agree with you 100%, you have to keep in mind of where this methodology probably stems from; the earlier days of Sub-D modeling. Back then, using triangles created awful artifacts or broken models, and n-gons would fail to render the entire model itself or cause crashing. My point is, the advice adhered firmly into the general rules of a lot of people, and those that were using it back then have handed down the adage of "quads only". So if you're using Maya 4 or 5, then yeah, no ngons or triangles on subd shit :P
The rule had merit back in the day, but now it is more of a non-issue, and a lot of people don't realize this and just listen to the advice.
Although, this could all just be the enchiladas that I ate for lunch talking. Not sitting to well right now...... awwwwwwwwwwwwwwwwwwwwww mannnnnnn.
your example is good except for one thing - your model is made up of multiple objects that share a texture - if they are seperate objects you only need a greyscale map and a color swatch to modulate the spec and get the same effect. the option to use rgb spec textures is great from an artists perspective and I prefer it if its available.
You probably don't want to sacrifice resolution where you can afford to have it just to match a lower resolution where its a technical necessity - if you are stuck using a 2k map on the body and a have the option of a 1k map on the face you probably don't want to drop the face down to 512 or 256 just to preserve uniform density.
But you want to be very careful about density across UV shells sharing a texture, or else you will never be able to get rid of seams - I ran into this problem re-texturing old assets (a horse) where someone had given the body and head the same density but the neck about half and there is nothing you can do - baking, cloning, 3d paint - to fix those seams (short of re-uving)
Zac, the funny thing with that image is, it doesn't actually look good. You're just providing an image and saying "X game did this, so it must be correct". But the image itself is a very clear and obvious example of what inconsistent texal densities can do. I would wager to say that the character in that screenshot would look BETTER if the face was the same density, simply because the inconsistencies there draw a lot of attention to how lowres the body looks. If its consistent there is no nagging feeling there, it all just fits together.
Jedah, I dont remeber anyone saying "Learning Blender is a waste of time", or am I missing something?
I mean a dumb thing would be to learn blender only, then hope to be up and running day one at a professional studio not using it in its pipeline (for very specific and justified reasons other than fanboyism hehe). Of course if the portfolio is steller it wouldnt really matter to get the job, but in order to keep it one would likely have to learn another app, thats just a reality.
Regarding this - I would be very interested to see more Blender myths not necessarily debunked, but at least put to the test. For instance, its modeling toolset. Of course die hard Blender-only users swear by it ; but I would love to see a skilled Blender artist perform a series of simple and complex modeling tasks on video in order to tell if, indeed, Blender is an efficient modeling environment. In this sense, Blender CG movies don't prove much - they show that the end result can be great, but they say nothing about what it took to get there!
So you seem to be sold regarding Blender being a faster modeler than Max. I totally trust you in the sense that it seem to fit your workflow and that's what matters. Could you be a bit more specific? Maybe pointing to videos showing what you mean. If it feels like too much of a derail you are free to PM me.
(I actually do want to switch my modeling app to something else than Max but so far not a single one topped it, after extensive testing. I keep trying out Blender regularly but some things are just backwards for no reason, and I cannot trust a software showing the obvious signs of stubborn project management ...)
Still a great thread!
I mainly hear complaints about blender being a waste on places like Turbosquid, but it doesn't really matter.
Right now the best thing to do if you want to look into Blender's modeling speed is to spectate/lurk Blendernation. Blender is currently undergoing a huge transition with Blender 2.5 and Bmesh and a lot of it's core modeling features will be changed within the next 3~9 months during the beta.
Also note that I said Blender was more of a learning experience for me. While Blender is nice it is missing some things which hurt it ( The way Blender handles smoothing groups/Edge smoothing is terrible. )
Its a tough decision, peoples eyes natural focus on faces, and they can take up to 30+% of the screen space in some of the cut scenes in Mass Effect, they probably figured being able to so extreme close ups was more important than a difference in density. And it looks a bit better in motion with the TV style dialog and mini cut scenes.
For a portfolio or still piece, its always better to have a consistent density.
I find it hilariously ironic that zacd has taken it upon himself to instruct per and earthquake in the terrible advice thread.
Anyway i always interpret the increased texel density on the head thing as a way to create a focal point with sharpness, and that raises a question:
I've seen uneven texel density have a super-powerful positive effect, especially in diffuse only work, as a way to put sharpness where it's desired and take the focus off the other bits. b1ll's diffuse work is a good example of this: http://benregimbal.com/lpsmall.html
How do you guys feel about this? Do you think even texel density is one of those general rules that should be broken if a very skilled artist feels they can do better by mixing it up, or a hard and fast directive to stick to?
I still think it's a solution to a limitation rather than a deliberate choice. We'd have to hear from Ben himself but I assume he would've made every island the same density if the budget allowed for it but since you can't scale everything else up to the face density you compromise the less important parts. I understand this feels like the "rule" because, well, pretty much every game has these kinds of limitations but as I see it ideally islands should have equal consistency.
I find it hilariously ironic that zacd has taken it upon himself to instruct per and earthquake in the terrible advice thread.?
I just figured since I heard advice here that is opposite to what I see in games, its worth at least discussing.
There was someone in the IRC that asked if he should give the eyes a higher density cause they are a focal point (and he had room), and no one had a real solid answer.
There is a pretty common misconception that if you can save a few tris by using alpha, its worth it. This however, is quite misguided.
1. If you're keeping the alpha'd bits on the same UV layout, you're going to DOUBLE the texture memory on a compressed diffuse texture. A 24 bit .dds is 1/2 the size of a 32 bit .dds
2. The increased shader complexity and need for alpha sorting is going to be much less efficient to render than a small amount of geometry, in virtually every modern game engine(may be different for portable systems).
3. Adding a 2nd smaller texture is still going to eat up memory, and also introduce a 2nd draw call for your mesh, again which will likely hurt performance more than a few extra tris.
So, moral of the story, do not use alpha unless it is absolutely necessary, and especially not to save a few tris.
Is that coming from Jolyon Myer's pdf where he was reducing 50-60% of a model's polygon count by using alpha-tested planes for cylinder caps?
I was always wondering if it would be worth it or not. On one hand polygons seem to be cheaper these days, but on the other hand it would take a fair number of them to get the same roundness on all those wheels (and there's 20 of them)
as for Joes "2." alpha sorting wouldn't be required in the case of just using alpha-tested stuff to replace tris. His remark about shader complexity still applies of course. You'd not want to add alpha-testing only because a tiny bit of a mesh would use it.
To disable backface culling for the entire thing if only a few polys need to be duplicated is wasteful. The majority of backfaces of thee object, that don't need this would hurt performance. For "batching" it would also be bad to mix objects that require this to be turned off and others not. However if you have a lot of stuff that requires it off, and all can be combined (think lots of window planes) then there is no need to duplicate manually as the hw rasterizer does it for free.
cryrid: Like with everything, it's important not to get too black/white on issues. Joe's language is clear. 60% of a model's triangle count is not "a few tris". Naturally you need to look at the specifics of each case.
My bad if I sounded black/white; I know that many game-art answers seem to fall back to 'it depends'. It's just that "a few" and even 50% are all so relative that it's hard to tell what is meant. Depending on the artist/engine/whatever 'a few' could be 16 tris or 1600. There's another example in the pdf where it's used on a more simple cylinder. Still knocked 50% off, but it was only around 32 triangles saved in the end which seems a lot more questionable. The train though seemed like a decent example of when the initial advice might be applied in a thoughtful way, so I thought I'd post it.
A former lead programmer of mine told me a rule of thumb for using alpha (this was for our engine at the time and may not apply for everything) - Transparency will cause a model to be rendered twice (for backfaces), so if you can cut the vert count at least in half by painting the holes in the model, then use alpha, otherwise, just model the holes. Example - If you have a 1000 vert model with alpha, the engine will actually be drawing 2000 verts. So, if modeling the holes will not increase the vert count above 2000, then its more efficient to model them and not use alpha.
However, an alpha channel will also double the size of a DDS, so thats another consideration.
Cyrid: Yeah like per says, my main point is that using alpha to save a relatively insignificant % is a poor idea, if you can make huge savings, then its worth taking serious consideration into the pros/cons.
CB: I actually was going to add a #4. but figured i had made my point. anyway;
4. Alpha blend is slow, and requires glitchy sorting. Alpha test looks like crap, and can't be anti-aliased without special features.
Replies
Yeah the sun is brighter, but the sky has just as much influence because of its size. Stand in the shade, anything that isn't touched by the sun or is shadowed, these areas are light by the sky.
The sun isn't always around is it? But the sky is.
Ok, im sorry for derailing this great thread.
Add to it the technical dimension that EQ talked about and you have a super complex skill set that didn't just fall out of the sky. I could almost go along that art is easier for some people (there is a long list of why that could be that doesn't relate to some mysterious art glad that is yet unidentified) but even then they still have to learn a mother truckload of technical skill to make their art functional. Which for most artists is counter to how their brain normally operates.
Someone calling an artist talented doesn't necessarily piss me off, it depends on the way they say it. When someone starts throwing around gift or gifted then that starts to grate on my nerves instantly.
Edit: This is getting slightly offtopic, sorry for that.
your example is good except for one thing - your model is made up of multiple objects that share a texture - if they are seperate objects you only need a greyscale map and a color swatch to modulate the spec and get the same effect. the option to use rgb spec textures is great from an artists perspective and I prefer it if its available.
I found that learning how to use Blender greatly improved the way I think about topology. Learning how to model a "different" way. Before I started Blender I was simply sticking to my student licensed 3DS Max, and a lot of times it felt like I was more of a 3DS Max user than an actual CG modeler. Blender was so different that now that I know it I find that I can switch to any 3D package with ease.
Plus Blender's shortcuts are hard to learn but invaluable when you do as it allows for some very quick modeling. I still primarily use 3DS Max because of it's status as industry standard, but Blender's usefulness as a speed modeler is not to be scoffed at.
Plus, Blender is entirely capable of creating top stuff. [ame]
I scale up bits and pieces if I've packed it in and there's a few little percent I can squeeze out of the remaining space. Or if some part of the model is so small and I know I want a certain amount of pixels on it for a particular detail. The difference in texel density shouldn't be noticeable if it's packed sufficiently well at 1:1 texel ratio beforehand and it gives you a bit more texture space to play with.
And I'll tweak island UVs to have straighter/aligned edges. Usually the stretching isn't noticeable and it's a lot easier to texture. Makes them easier to pack, too.
You can get away with skewing the density a fair bit. It's when it gets that some parts have twice the texels of other bits that you're taking the piss and your UVs suck.
It is very important to keep in mind the math that goes into scaling uvs. For instance
125x125% = 1.5 pixels for every 1
150x150% = 2 pixels for every 1
200x200% = 4 pixels for every 1
So even scaling something 125% can make a rather large difference. I think people make the mistake of thinking that when you scale something 125%, its only 1.25:1, instead of 1.5:1, so this can easily get out of hand.
Scaling things +/- 10-15% or so can be fine, but when you get higher/lower than that you need to be very careful.
Also, if you're scaling up AND down to fit everything in, you get your consistency out of whack really fast. Scale one thing down 90%, scale another up 115%, and all of sudden one piece has pixels 1.5 times bigger than the other, not good.
On the topic of talent and whatnot.
http://blogs.wnyc.org/radiolab/2010/07/26/secrets-of-success/?utm_source=feedburner&utm_medium=%24%7Bfeed%7D&utm_campaign=Feed%3A+%24%7Bradiolab%7D+%28%24%7BRadiolab%7D%29
I love the 10000 hour rule!
I mean a dumb thing would be to learn blender only, then hope to be up and running day one at a professional studio not using it in its pipeline (for very specific and justified reasons other than fanboyism hehe). Of course if the portfolio is steller it wouldnt really matter to get the job, but in order to keep it one would likely have to learn another app, thats just a reality.
Regarding this - I would be very interested to see more Blender myths not necessarily debunked, but at least put to the test. For instance, its modeling toolset. Of course die hard Blender-only users swear by it ; but I would love to see a skilled Blender artist perform a series of simple and complex modeling tasks on video in order to tell if, indeed, Blender is an efficient modeling environment. In this sense, Blender CG movies don't prove much - they show that the end result can be great, but they say nothing about what it took to get there!
So you seem to be sold regarding Blender being a faster modeler than Max. I totally trust you in the sense that it seem to fit your workflow and that's what matters. Could you be a bit more specific? Maybe pointing to videos showing what you mean. If it feels like too much of a derail you are free to PM me.
(I actually do want to switch my modeling app to something else than Max but so far not a single one topped it, after extensive testing. I keep trying out Blender regularly but some things are just backwards for no reason, and I cannot trust a software showing the obvious signs of stubborn project management ...)
Still a great thread!
Yes. Consequently, if you get a mirror and set it on the ground there's going to be a hell of a lot of soft blue and a super bright sun reflection.
But skin isn't near reflective enough to send any of that blue color torward the eye, only the very brightest lights are going to bounce off and make it all the way to us -- artificial lights near the surface, and the sun are bright enough to cause highlights on the skin.
And like per said, specular defines the object. Throw a grey spec on a grey object and fill the scene with bright green lights, you will get green highlights.
This has limited scope, blender while being extremely capable has a limited install base in studios compared to max or maya. If you have the choice id say learn a big name app, if your working in house you'll probably not be using blender so after getting used to blenders rather unique way of doing things you need to learn the other way too..which is a pain (I had to). You can get an edu license of maya of max for free at the moment so id push newbies in that direction myself. not because they are more capable but purely because they are more widely adopted in production.
That's hardly a myth. On high end hardware the triangle budget is huge of course, but even so, there's a limit.
And on hardware like the DS where there's a hard limit of 2000 per frame, then the triangles are not inexpensive at all. It just depends on the platform.
While I agree with you 100%, you have to keep in mind of where this methodology probably stems from; the earlier days of Sub-D modeling. Back then, using triangles created awful artifacts or broken models, and n-gons would fail to render the entire model itself or cause crashing. My point is, the advice adhered firmly into the general rules of a lot of people, and those that were using it back then have handed down the adage of "quads only". So if you're using Maya 4 or 5, then yeah, no ngons or triangles on subd shit :P
The rule had merit back in the day, but now it is more of a non-issue, and a lot of people don't realize this and just listen to the advice.
Although, this could all just be the enchiladas that I ate for lunch talking. Not sitting to well right now...... awwwwwwwwwwwwwwwwwwwwww mannnnnnn.
because nobody would want color variation, right?
Worrying too much about keeping pixel density consistent is also bad, a lot of games use different densities for the head.
You probably don't want to sacrifice resolution where you can afford to have it just to match a lower resolution where its a technical necessity - if you are stuck using a 2k map on the body and a have the option of a 1k map on the face you probably don't want to drop the face down to 512 or 256 just to preserve uniform density.
But you want to be very careful about density across UV shells sharing a texture, or else you will never be able to get rid of seams - I ran into this problem re-texturing old assets (a horse) where someone had given the body and head the same density but the neck about half and there is nothing you can do - baking, cloning, 3d paint - to fix those seams (short of re-uving)
and it drives me crazy!
I mainly hear complaints about blender being a waste on places like Turbosquid, but it doesn't really matter.
Right now the best thing to do if you want to look into Blender's modeling speed is to spectate/lurk Blendernation. Blender is currently undergoing a huge transition with Blender 2.5 and Bmesh and a lot of it's core modeling features will be changed within the next 3~9 months during the beta.
Also note that I said Blender was more of a learning experience for me. While Blender is nice it is missing some things which hurt it ( The way Blender handles smoothing groups/Edge smoothing is terrible. )
For a portfolio or still piece, its always better to have a consistent density.
Jack's however was terrible in ME2 even when moving.
Anyway i always interpret the increased texel density on the head thing as a way to create a focal point with sharpness, and that raises a question:
I've seen uneven texel density have a super-powerful positive effect, especially in diffuse only work, as a way to put sharpness where it's desired and take the focus off the other bits. b1ll's diffuse work is a good example of this:
http://benregimbal.com/lpsmall.html
How do you guys feel about this? Do you think even texel density is one of those general rules that should be broken if a very skilled artist feels they can do better by mixing it up, or a hard and fast directive to stick to?
I just figured since I heard advice here that is opposite to what I see in games, its worth at least discussing.
There was someone in the IRC that asked if he should give the eyes a higher density cause they are a focal point (and he had room), and no one had a real solid answer.
There is a pretty common misconception that if you can save a few tris by using alpha, its worth it. This however, is quite misguided.
1. If you're keeping the alpha'd bits on the same UV layout, you're going to DOUBLE the texture memory on a compressed diffuse texture. A 24 bit .dds is 1/2 the size of a 32 bit .dds
2. The increased shader complexity and need for alpha sorting is going to be much less efficient to render than a small amount of geometry, in virtually every modern game engine(may be different for portable systems).
3. Adding a 2nd smaller texture is still going to eat up memory, and also introduce a 2nd draw call for your mesh, again which will likely hurt performance more than a few extra tris.
So, moral of the story, do not use alpha unless it is absolutely necessary, and especially not to save a few tris.
I was always wondering if it would be worth it or not. On one hand polygons seem to be cheaper these days, but on the other hand it would take a fair number of them to get the same roundness on all those wheels (and there's 20 of them)
To disable backface culling for the entire thing if only a few polys need to be duplicated is wasteful. The majority of backfaces of thee object, that don't need this would hurt performance. For "batching" it would also be bad to mix objects that require this to be turned off and others not. However if you have a lot of stuff that requires it off, and all can be combined (think lots of window planes) then there is no need to duplicate manually as the hw rasterizer does it for free.
However, an alpha channel will also double the size of a DDS, so thats another consideration.
CB: I actually was going to add a #4. but figured i had made my point. anyway;
4. Alpha blend is slow, and requires glitchy sorting. Alpha test looks like crap, and can't be anti-aliased without special features.