The 3D Thread

[Rossco]

Well he did say it was cartoon style but rossco, as a starter I'll give you some advise to take on board (no pun lol!!) for next time...

- Overall the shape does work (though a bit toooo basic) but the main problem is the door area, it doesn't follow the shape of the body that well. This is most apparent in the front perspective view.

- For cartoon style the texture is not too bad, it's good that you have broken up the window and door with trims but the inside area has a really bad choice of texture (very cliche) and isn't broken up at all. If there was a trim on the inside also it would help break it up. I'd use a more basic metal material for that also.

Not bad for a start though, the front is quite a nice shape and the 'landing ski's' (not sure what they are called tbh) do the job, I like how they are curled.

If you read back in the thread a bit, you'll see that Rossco first approached a 3D program under 6 weeks ago. He's literally just started a degree in this subject area.

There's being helpful, constructively critical, then there's being blunt and dismissive. He's just started so Capone and I, without needing any kind of communication about it, decided to go down the route of tact and diplomacy.

Yes, the model is basic, the texturing is simplistic, and it's subdivided to buggery - and there are probably all manner of other problems. What you don't realise though Cheeko is Rossco has almost certainly been taught a set of skills, and then been told to utilise them to complete a set brief.

We've all made poor models or basic models when we started. My first ever 3D model, a mech made in Bryce from a tutorial in 3DWorld magazine was similarly basic. That doesn't mean it wasn't a valuable learning experience. There's no need to jump up and down on his beginner-standard work.

See that's the kind of criticism I'm looking for so thanks for that guys but yeh you're completely right about the door area, I don't really know why I kept it that shape as it obviously juts right out and looks daft! Especially from the front as you said.

What do you mean about the trim on the inside, like something along the edges of where the wall meets the floor instead of being curved? I do quite fancy keeping this model going to see if I can improve it anymore or maybe I should start a completely different one and try and model an actual real life one.

Asura, you said it's subdivided too much, is that a bad thing then I take it? What techniques should I use to avoid this?

[Asura]

Asura, you said it's subdivided too much, is that a bad thing then I take it? What techniques should I use to avoid this?

All I mean by that is the polycount looks very high. Do you know how many triangles it is?

It always depends what you're using it for, of course. If it's just for an animation then the sky is pretty much the limit (at least, the limit is the max your computer can handle to get rendered within the timeframe).

With games however, polycounts are much lower. You have to use things like textures, normal maps and smoothing groups in order to reduce the count.

[Rossco]

Yeh good point, I think I said before but the lecturers haven't mentioned once to watch the poly count but it definitely makes sense to pay attention to it now as ideally I want to get into games down the road so it's obviously a big deal with that.

So funny though, rendering this on my PC at home is ridiculously faster than the uni PC's which is quite shocking really. It is about 5-6 times faster on my machine at least for most scenes!

Now you mentioned triangles but I thought you were to stay away from triangles so why do you ask that?

I done a wee check and it says total faces is 25,021, 96 of them are 3 vertices and the rest are 4 with a handful more than 4.

[Rossco]

I don't think I mentioned but I realised a few hours after I handed this in on Friday that I put the wrong file on the DVD, it was one encoded at 12fps and not 30! Assignment isn't due until tomorrow though so I can still hand in the correct one which is lucky! The links I put above are the wrong version so I've uploaded it again now! It's still low quality though even though the file I uploaded looks way better normally. Confusing!

If you click the link and look at the similar videos it lists you might see some others from my class if it says they've been uploaded in the last few days and it will be around a similar environment shape, so you can see the difference in quality. Not saying mine is amazing but much better than the majority out there so I'm happy! Would be unfair to link to them specifically though I think although I did laugh at a few.

[Asura]

Now you mentioned triangles but I thought you were to stay away from triangles so why do you ask that?

"The traditional wisdom" would state that you should stay away from using triangles, yes, but sadly real life is more complicated.

First off, everything onscreen is composed of texels (textured triangles) anyway. Even if you have four-sided face, your renderer will split this into two triangles at render-time. If you had a six-sided face, or eight-sided one, your renderer would likewise split that too into triangles.

The problem comes because you can't predict necessarily how a renderer handles that splitting. Different renderers do it differently, and the same renderer might split the same poly differently depending on the angle.

Four-sided polys are ok, because how they split is predictable (the order in which the points are given to the renderer dictates it, and most modelling packages allow you to change this manually, "turning edges" or similar).

Whether or not triangles are ok depends on certain factors. On a flat surface, if they're coplanar with something else, you can get away with some triangles. Also, modelling is a visual medium; provided you're not working with simulation or architecture, if it looks right then it's right; subsequently if it saves you a lot of polys, you can stick some triangles in here and there.

Remember though triangles cause some problems too. Sometimes they don't illuminate correctly, which can cause problems with specular highlights (it causes problems with diffuse lighting too, just the effects aren't as acute). Also, if you're using a skeleton to deform a character for animation, triangles can deform in weird ways. Lastly, if you subdivide/tessalate your polys to make things smoother (usually when you're making a high poly model for normal mapping) triangles will make nicks, scrapes and uneven bits.

Conversely, if you're modelling something like my Advance Wars models, you can use a lot of triangles because they're extremely low-poly and they're not going to be deformed, subdivided or even lit.

I done a wee check and it says total faces is 25,021, 96 of them are 3 vertices and the rest are 4 with a handful more than 4.

Generally speaking, modellers list their triangle counts as opposed to polycounts. You should try to specify this too. For example, my GM Custom model is ~4000 triangles. If I said 4000 polys, I could potentially mean as much as 8000 triangles, or as little as 4001. Using triangles is a lot more informative. Often if I look at someone's portfolio and I read "3000 polys" I unfortunately always wonder if that means "6000 triangles". A difference like that in polycount can make an impressive piece of work into a very poor one.

[Asura]

Thanks for the tip.

Last night i spent a good while playing wiht just the unwrap feature (i know i could have used cilindrical projection for this, but i wanted to practice properly.

Tonight... Mudbox!

Reminds me of something I worked on a while back for an Unreal 3 mod:

They're pretty low-poly, though I can't remember the actual count. I think they had the bottom and 16 sides, so that'd be 48 triangles.

[Ste Pickford]

First off, everything onscreen is composed of texels (textured triangles) anyway.

Texels aren't textured triangles. They're the individual pixels within your texture bitmap.

If you draw a 64x64 pixel bitmap graphic, and use it as a texture on a 3D surface , properly speaking that's 64x64 texels.

The reason for the separate word is because once a textured triangle is rendered in 3D, the word pixel is used to refer to pixels on the screen / viewport (the size of the window being rendered - say 640 x 480), and the word texel refers to an individual pixel of the texture bitmap (if that makes sense), which is now at a size & shape completely independent of the pixels making up the rendered image.

If you zoomed up close to a textured triangle in 3D, then you would be seeing an individual texel (of the texture) being rendered into many pixels (of the screen).

Equally when you zoom very far away from a textured triangle you'll find a single pixel (of the screen) is being rendered based on the colour of several texels (of the texture).

I understand that people are taught to model using quads and stay away from triangles. I don't really know about that, but in terms of terminology in games, you might as well consider the word polygon to mean triangle. When people talk about poly counts in models, or how many polys an engine can handle per second, etc., they almost always mean triangles.

[Ste Pickford]

One thing about the helicopter that stands out as weak is the massive mis-match in poly density between the body of the 'copter and the tubular piping rims around the door frames.

If the spec is to do a low poly model, doing some densely curved cylinder mesh like this is an instant fail. If the spec isn't to go for low poly, then having such a low poly body in comparison to the piping would also suggest poor modelling skills.

I always try to keep the poly density fairly even across the whole model. This helps with lighting, as well as gets you the best value per poly, particularly when making low poly models.

If you find there's one a bit of model that needs to be massively denser than others, then there's a good chance you're building it wrong, or you've not thought it through properly.

Very often anything that initially seems to need a lot of densely packed polys might be much better being done in the texture - ie painted on (which would suit the piping around the doorway and window frame of your 'copter).

I spend a fair amount of time planning models out on paper first - just little sketches to work out the breakdown of primitive shapes, where the bulk of polys are going to be needed, where I can save a few and where I can afford to spend a few more etc - before I do a single thing in the 3D editor. This can be a big help.

Also, when I've been recruiting, I'm always very suspicious of models that are well rendered in Max, with lots of reflections and shadows and stuff. I always feel this is to disguise weaknesses in the model. If I'm judging the abilities of the modeller, then just give me the clearest views of the model itself, don't try to dazzle me with rendering effects. The modeller doesn't do rendering. I'm not arsed about rendering. It also makes me think the modeller is someone who spends too much time on the wrong things. If you've got 2 days to make a model, I want to see 2 days of modelling or texturing work. Not a day and a half of modelling, and half a day messing with lighting and rendering effects to show the model off.

Nicky's Max views of his barrel model are perfect for presentation, as I can see everything about how the model is built, how it's textured etc., and from that can judge exactly how it will look in a particular game engine, or with certain lighting conditions etc. The helicopter render with raytraced shadows and reflections is much harder to judge from a modelling perspective.

[Number 28]

Not sure I agree with that Ste. The piping needs to look nice and smooth, hence more polys. You put more polys where you need more detail, or in this case, a nice smooth finish. Besides, the views above are deceptive, because the polys on the helicopter 'shell' have been smoothed. There'd be double the polys on the outer shell if he subdivided* to get the same result (*not sure about the correct Lightwave terminology).

[Ste Pickford]

Not sure I agree with that Ste. The piping needs to look nice and smooth, hence more polys. You put more polys where you need more detail, or in this case, a nice smooth finish.

Yeah, that's true. I suppose what I mean is that if I was building something that needed dramatically more faces or vertices in a one small part, I would seriously question whether I'd designed the model correctly - or smartly.

You need more faces for the piping to get the exact effect required by that particular model, but that suggests to me that perhaps poor decisions have been made when conceiving the model in the first place.

Do you really need such smooth piping? Rather than how do I make smooth piping with fewer polys.

[Number 28]

I guess it does look a little dense...like, solid white. I usually take some rings out where the pipe isn't bending... if that makes sense.

[Asura]

I understand that people are taught to model using quads and stay away from triangles. I don't really know about that, but in terms of terminology in games, you might as well consider the word polygon to mean triangle. When people talk about poly counts in models, or how many polys an engine can handle per second, etc., they almost always mean triangles.

You're of course right when you say that the terms polygon and triangle are interchangeable. The only problem is MAX lists both triangle counts and "facecounts". I've known people to use the "facecount" as the "polygon count", either on purpose to mask their true polycount or simply because they don't understand/care about the difference. Either way, it's wrong.

Not sure I agree with that Ste. The piping needs to look nice and smooth, hence more polys. You put more polys where you need more detail, or in this case, a nice smooth finish. Besides, the views above are deceptive, because the polys on the helicopter 'shell' have been smoothed. There'd be double the polys on the outer shell if he subdivided* to get the same result (*not sure about the correct Lightwave terminology).

I think you're both right, for different reasons. As was said earlier by Cheeko, research is the key. If, for example, the model was being used for an Ikaruga-style 2.5d shooter, then potentially the skiffs/skis/landing thingummy-bobs could be merely a couple of polys with alpha-blended textures applied. Alternatively, if it was for Rainbow 6 you're obviously going to need to take more care.

For example, in my GM Custom model, the mobile suit's backpack is slightly higher resolution in terms of texturing than the rest of the model. It's because I took the liberty in my brief of assuming the game would be played from a perspective similar to Lost Planet/Mechassault, and the player would spend the bulk of the time staring at the mobile suit's back. Stuff like Ryo Hazuki's out-of-focus tiger on the back of his jacket annoyed me no end in Shenmue; it was a terrible, murky texture and was displayed onscreen throughout almost the entire game.

Also, when I've been recruiting, I'm always very suspicious of models that are well rendered in Max, with lots of reflections and shadows and stuff. I always feel this is to disguise weaknesses in the model. If I'm judging the abilities of the modeller, then just give me the clearest views of the model itself, don't try to dazzle me with rendering effects. The modeller doesn't do rendering. I'm not arsed about rendering. It also makes me think the modeller is someone who spends too much time on the wrong things. If you've got 2 days to make a model, I want to see 2 days of modelling or texturing work. Not a day and a half of modelling, and half a day messing with lighting and rendering effects to show the model off.

This was something else that was a bit frustrating at the Eurogamer Expo. Several of the recruiters/lead artists at the place pointed to bits of my portfolio, stating that they wanted "the finished article". They then pointed to stuff like my Ico environment windmill and the fighter plane, stating they were "the finished article" but my GM Custom model wasn't.

Both those projects are nowhere near as good as the GM Custom model. The Ico thing for starters is really just an art project.

At first I was confused about it, but after hearing it several more times, I spoke to someone about it who'd made a similar comment, and was told why - the Ico scene and the plane are rendered using advanced lighting techniques, whereas the GM Custom model is rendered as it would appear in-game (using basic lighting).

One of the blokes who saw my GM model even said he was a Gundam fan, and thought I should model some backgrounds with explosions and whatnot in, and pose the GM model in them, in a dramatic fashion. That was the point I when, as I said in my previous post, I realised that these people were only human too, with wildly differing preferences for what they want from candidates. I've been told since uni not to try to blind recruiters with fancy rendering techniques and dramatic scenes when you just intend to model, but the whole experience at the Expo made me wonder if to stand out from the crowd, do you really need fancy renders?

Then, you'd think "well, you could supply those with other, more informative ones" but some of the booths were saying they'd prefer you just had THREE images in your portfolio, no more, no less - and that three images were all they would look at before moving on to another candidate.

All these factors were what contributed to me saying "fuck it" to these small projects and setting out on the Gears thing Wroth and I are doing now.

[Rossco]

I love coming on to read discussions like this. Its great hearing peoples opinions and ideas!

Asura when you referred to triangles and that everyone when stating polygons would mean triangles, do you mean that with my 25,000 poly'd helicopter, I should be saying it has 50,000 triangles since that's how many it would be if I turned the quads to triangles?

But regarding the piping Ste mentioned, it's definitely a weak point but I think that is my fault due the the weird shape of the door opening anyway! I didn't really bother to look at it much or change it, when I made the opening I just went with the first shape that appeared when selecting polygons I think. If it was more square like with straight edges I could get rid of the majority of the rings all the way along the flat sections and just keep alot at the corners. So that's something to focus on it future, just lazy modelling I guess. Thinking about it, the pipe is probably 24 sided too which is way more than is needed for such a small detail. But I did always plan on doing the up close shot with the camera right next to the piping reflecting the world around it so wouldn't have worked so well with less polygons as the reflection wouldn't look so smooth.

I partly agree with the idea of not putting lots of fancy effects on when rendering. I still think you have to show the model off, especially just now since this is just for uni and we haven't been given any limits so I will produce the full animation to the highest quality I can. Remember it is Computer Animation I'm doing and not just modelling so need to work on everything for the time being. I didn't bother with ridiculously high amounts of AA though or Radiosity either as that wouldn't be used like you say in the real world for video games.

Anyway, I have that class again tomorrow and I'm hoping the lecturer has marked them but I think that's wishful thinking considering he has about 40 of them to go through. I just hope he lists the marks highest first and have my name there, it's a great feeling when that happens!

I've started a model for a Microwave, hardly done anything but somehow I fucked something up and some polygons have been deleted next to the opening as you can see in the second picture and I don't know how to get them back! The Make Polygon button just makes an arse of it when I try!

Can't upload them to Imageshack today for some reason so had to use the attatchments here instead.

[SuperCat]

Wow, there's allot of debate here. Question though, does anyone posting here actually work as an artist?

[Number 28]

I have, but i'm pretty sure Ste Pickford has never worked in games, the charlatan.

[Asura]

Wow, there's allot of debate here. Question though, does anyone posting here actually work as an artist?

A couple of people do. I don't (I'm currently employed as a 2D graphic designer).

[c-cat114]

Hmm I don't know if I've ever posted in this thread but, if I have then yes, I work and have worked as a 3d artist (currently teach at a college computer arts and animation, formerly worked in a variety of games houses until I got a mixture of fed up, bored and made redundant three times in six months.)

Anyhow, loving the artwork Asura and Rossco. Bodes ill for me if I ever did want to get back into the industry (which I don't, but I occasionally get the itch) that the standard is so high in A's portfolio. Roosco's stuff is also very good for the beginning efforts in a 3d package and hopefully the constructive criticism is more heard than the more traditional Mukslinging.

Different but related subject are there any Blenderheads on the forum? I've been a long standing Max user but let it lapse to Max 7 while doing other projects. I've got a 3d project on my mind and fancy teaching myself Blender instead as some of the output is of such an amazingly high quality. Partially I want to end up teaching it to my classes but some parts of it are so...specific that I'd love to pick the brains of someone who's been using it for a while. Boolean is the first head scratcher for me. Theres a union, division etc but seemingly no option to carve one object out of another, which is a little confusing. It's such a huge omission I assume it's in there somewhere, but can't find it myself so far...

[Asura]

Different but related subject are there any Blenderheads on the forum? I've been a long standing Max user but let it lapse to Max 7 while doing other projects. I've got a 3d project on my mind and fancy teaching myself Blender instead as some of the output is of such an amazingly high quality. Partially I want to end up teaching it to my classes but some parts of it are so...specific that I'd love to pick the brains of someone who's been using it for a while. Boolean is the first head scratcher for me. Theres a union, division etc but seemingly no option to carve one object out of another, which is a little confusing. It's such a huge omission I assume it's in there somewhere, but can't find it myself so far...

I think a large part of it is detaching what can be done from the quality of a package. I've seen amazing stuff done in Blender and Wings 3D, but poor stuff in MAX and Maya. Obviously it's always going to vary but people can be really creative; and post-process stuff in Photoshop can count for a lot (my work is post-processed as little as possible, often not at all).

As for seemingly obvious features being missing, it might not be you - the feature may just not be there. For years, XSI didn't have an "extrude along normal" function like MAX's. It had extrusion along the cardinal directions, or extruding each poly, but not an "extrude group".

[c-cat114]

I think a large part of it is detaching what can be done from the quality of a package.

Yeah, I don't think I explained myself too well on that one. I've seen some awful stuff come out of Max and I know it's the person using it rather than the package itself. Blender, years ago, seemed a hideous mess and the results were mediocre. I was mainly surprised that Blender could output stuff as good as the more professional and expensive packages blow for blow nowadays. In fact some of the internal guts of it seems more powerful in ways: the Node Editor parts seem fantastically powerful. I guess what I was trying to say was that of all of the packages out there - if I were wanting to play with a 3d package it would seem to be the more exiting one of the bunch...

But yeah: may well not be there. All I wanted to do was create a simple photo booth (well, all I wanted to do was have students learn the process of carving a box out of another and makig a simple cloth modifier...) and there seems to be no specifically easy way of doing it. In the end we did the project in Flash, but I do want to get Blender more used in the future.

**edit: take it back - and a totally idiotic schoolboy error from me. Forgot that the choice of operands 1 and 2 would change the Boolean so everytime I did a difference boolean it was subtracting the bigger from the smaller and skewing the result. Picked in the right order gave me exactly the object I was expecting. Man I've been on Photoshop too long...

[Anaardvark]

I'm genuinely scared of posting anything on here now

Bearing in mind I've only been in my course for 6 weeks, I'm finding Max a bit hardcore to say the least. It's just knowing where and when to use all the different tools and use them correctly. We've moved from creating a clamp and animating it in a non-mechanical way, straight on to character modeling using a virtual studio.

If we've done all this in 6 weeks, god knows what we'll know by the end of the year.

[Asura]

I'm genuinely scared of posting anything on here now

Bearing in mind I've only been in my course for 6 weeks, I'm finding Max a bit hardcore to say the least. It's just knowing where and when to use all the different tools and use them correctly. We've moved from creating a clamp and animating it in a non-mechanical way, straight on to character modeling using a virtual studio.

If we've done all this in 6 weeks, god knows what we'll know by the end of the year.

I wouldn't worry. You'll get bombarded with a LOT of information, but most of next year will probably be about the use of that info, rather than giving you more.

Like for example, in the first three semesters at Teesside, they teach you the basics of animation, which is technical - but then the remaining animation modules involve you watching movies of stuff like Wallace and Gromit, Pixar, Disney etc. and learning how their animation works on an artistic level. This is easier in some respects, and more difficult in others - it's certainly more relaxing, but you need the groundwork in animation theory to have the skills to recognise what you're seeing.

[Rossco]

I really am not interested in character modelling or animation but for an assignment we have now, I have to model a human from top to bottom so here is my days progress so far using splines.

Since it's my first attempt at this it will probably look nothing like the reference image (me). Plus, the images don't line up properly and can't seem to fix them no matter how long I fiddle about with the size and position so it shouldn't look too much like me. But hopefully with some half decent UV mapping skills, I can map my face on it. I have no idea how hard that is though, so may end up a disaster but only time will tell! Will be quite cool though if I can be finished with something that vaguely resembles me in 3D!

[Capwn]

Front view is looking odd. Too 'square'. Side is looking nice though.

Sorry, had to do this with mouse but hopefully it helps...the important thing i'm trying to get across is the difference of the inner green lines. Also the shape of the upper head

[Number 28]

Can you not use polys instead of splines? Never got the hang of those things...

[Rossco]

Yeah very good points there, dunno why my forehead splines are all dodgy, I'll go back and fix them, not sure what to do about the jaw/cheek section though, it's a mess I think and just keep tinkering with it.

Think I'll get stuck into it again now for a few hours anyway and come back with a complete head soon for criticism.

28, alot of people are box modelling their heads for this assignment but I don't like that as I think it always ends up looking like the same generic head as y ou do the same set of operations on the same polys time and time again so thought I'd try the old spline method and see how it goes. But as I said before, I'm not too bothered anyway as character modelling is not something I plan to pursue in the future.

[Rossco]

Yeah so I finished doing my splines, and then patched it all together, and mirrored it and this is the result so far.

Looks kinda creepy! The mouth seems to stick out quite a lot, possibly the lower lip is just a bit too big and far forward. And also the forehead seems a bit pinched together in the middle but that should be fixed easily enough. The nose is a bit thin I think also down the front of it so maybe thicken that up a bit. Anything else that looks out of place majorly? I did fix the forehead spline alignment as you see in the front view so they are much straighter now thanks!

[Rossco]

More progress:

I'm happy with most of it so far, there is still a bit of pinching around the forehead that just won't go away right now. Spent ages doing eyes, nose and mouth, really tidied them up and I'm happy with most of them. Ears however are proving to be right tricky bastards! Not finished with them yet though, should manage them fine. I'm really spending way too long on this as the detail here is not required on the head for this assessment, but the full body. But I'm enjoying it and learning all the time so it's all good.

[Rossco]

Okay so I really shouldn't and can't spent anymore time on the head for this model so this is it. I spent ages on the ear and still not totally happy with it. But for my first head attempt, I'm pretty happy overall.

Now onto the body tomorrow.

[Ran]

Yeah thats coming on nicely I think. If you do any more head tweaking i'd look at bringing out the cheek area, its too drawn in giving the impression that they're being sucked in at the moment. Looks really good in the profile view.

[Rossco]

Yeah cheers, I kinda see the sucked in cheeks now! Might change that, not gonna spend much more time on it though, got plenty more stuff to model first! Plus I still have to texture the head which I am not looking forward to in the slightest. I so haven't got the hang of UV maps yet but I will persevere.