Texture Size, Pixel Counts, Video Memory, and File Formats

they "dont" after every save, its because your image editor is not working with a compressed image, so its like first time quality degradation (or 2x if it was already a jpg or 3x ect)

but save an image as a jpg, close it, reload the image and save it ... do that about 10x @ 50% quality and your image will look like some bad NES graphic really quick, course the higher the "quality" the longer you will be able to get away with it, but even at 100% quality after 10x it starts to fuzz, loose edges and get grainy

damn fourms ate all the attachments, but at one time i did have an image up showing just that

have you ever tried that?

btw, colour profiles might be a pro/con

dont you think?

(puts on his anti slander hat)

i just said i did

and no

heres one i did for an "art" peice in sl

the original image was a uncompressed 24 bit bitmap, after about 100 passes tru jpg compression (using scripted image magic) and 1 cartoon filter in the gimp this is what you get

nope you just said you didnt

oops the hat just fell off

are you retarded or something



what the @#*! did you think THAT was?? a image of my bird ?

done with you

i cant see the original image

neither how many times its been saved over and over

neither a pixel by pixel comparison

then do it yourself

remember just cause you saved an image in whatever shop a billion times wont cause it cause your not loading pre compressed images

load save unload reload and save, do it at 50% so its quicker

but at this point your dont even care, doubt you did anyway, you just want to bicker

now what do i have to do?

save, close, reopen etc?

ps

is that ok for a test image?

http://www.colour-science.com/quality%20test%20tools/test%20files/Reference%20Print%20printer%202362x3543pixel.jpg

pps

after five iterations.. is it the jpeg , really?

Johsua, let me say two things. First of all, the answer to your latest question is no, I would not consider the image you linked to be all that well suited for testing purposes. It's a print-size image, not something that would ever be used as a texture. Since the topic here is texturing, it would only make sense to test with the kind of imagery that is used for that purpose. I'd suggest you perform your tests with something 256x256 or smaller.

Here's why. One of the things that makes JPEG particularly bad for textures, while at the same time relatively OK (but not ideal) for photography, is the fact that textures are small, and photos are huge. The larger an image is, the less noticeable the impact of compression artifacting. With a large image, each individual artifact will be relatively smaller in comparison with the whole canvas, so they're individually harder to spot.

That said, if you really want to do a pixel by pixel analysis of an image that size, be my guest. The same kind of degredation will occur with each re-comression, regardless of size. It'll just be a lot more work to complete your analysis in any meaningful way with such a big picture.



Second, if you must continue talking about this subject, please take the discussion to a new thread. It does no one any good to continue it here. Between your earlier attempt to introduce topics that were not at all relevant, your insistence that relevant facts are untrue, and your intermittent claims that things people have said are somehow the exact opposite of what they actually wrote, all you've accomplished so far is the potential confusion of those who would otherwise benefit from learning the material this thread has to offer.

Perhaps sabotage was your goal right from the start, or perhaps there's a legitimate, perfectly innocent misunderstanding in play. I have no idea. Either way, it's not right that all this useless back-and-forth be so etched in stone in a sticky.

Here's what would be the right thing to do. If you want to do testing, go right ahead, and if you want to discuss your findings, start a new thread for that purpose. If anything you discover warrants alteration or update of anything in the OP, I promise I'll make the appropriate changes, and give you due credit. Short of that, this discussion should end here and now, so the thread can be allowed to get back on topic.

Thanks in advance for your cooperation.

uh, the image above says "you caught me. sorry"

"is the fact that textures are small, and photos are huge" 2:0

"I have no idea. Either way, it's not right that all this useless back-and-forth be so etched in stone in a sticky."

backanforth useless. etched, sticky

Hi Chosen:

I see this thread was started a couple years ago.
So if you're still around.......

Thank you!!

That *really* helped me. Especially since I'm analog by nature.

/me giggles

I'm learning so much.

There is *so* much to learn.

Thanks again!

Please help.. I am new to photoshop... know it not to bad... is there any way to use a texture or fabric I already have made and just add it to the templates? I have searched and searched and no clue if you can?

And how does one draw a straight line with a vector tool and have it fill in? OMG ripping my hair out here. Step by step would be awesome... ty in advance))

Minx Maggs

Minx, please post your questions in a new thread, since they've nothing whatsoever to do with the topic of this one. I promise I'll answer your questions (and others probably will too), but it's super important to keep the stickies on topic, so I won't do it here. Start a new thread. Thanks.

Chosen,

Are you a teacher in RL? You explain things so well and make things very easy to understand.

you did a great job expressing many points on the file formats.

i like png. mostly cause i find working in alphas a pain compared to just save as... png. no matter what i use jpg,targa, heck even png i find the most degradation isn't in the lossyness of the compression but rather by the drop resampling the ppi. i like to work in 1024 @150 ppi and then save the finished item AS png. Then i open the png file and resample to 512 @ 72 ppi. its not to bad. depending on the detail ie tatoo, clothing, wall i may have to apply some unsharpen filter. then resave as png(lossless). Any thoughts?

jpegs compress to save space. the way they do this is by averaging pixels that are by other pixels. if a bunch of pixels make up a grey patch then the compression simply replaces the many colours that made up the detail in that grey patch with grey. the next time you open the image up and save it .. it now takes that grey patch and compares it to the pixels near it and determines that the overall is a blue grey... and replaces not only the many pixels that made up the adjacent blue grey colours and the grey patch and reasigns the whole patch and adjacent pixels to bluegrey. You have now lost all detail in your image. you can open the image up as many times as you want ... its only when you save AS JPEG that the whole formula of averaging colour starts up again. you can save a jpeg that you found as a tiff or ps or any lossless file and then work on it as many times as you want as long as your not calling upon the jpeg compression.

As I, and others, have written about extensively all over this forum, you're shooting yourself in the foot in ways you can't even yet comprehend by not learning to use alphas. It might not seem like it, but the fact is by working only with WYSIWYG transparency, you're making a lot of things take far longer than they should, you're sacrificing at least 99% of the control you'd otherwise have, and you're even making certain things impossible to do at all.

The only thing that is a "pain" about alpha mapping is getting the proper mindset about it into your head for the first time. But once you've done that, it couldn't be easier. WYSIWYG is actually far more difficult, in many respects. The vast majority of people don't tend to want to believe that at first, so you're certainly not alone in your current thinking. But without exception, every single person I've ever taught to make the switch has been pretty amazed to discover just how severe their previous self-inflicted handicap actually had been.

Alpha mapping applies to soooooo many other things besides just transparency for SL. It's the very basis of how digital graphics works. I'd highly recommend you stop thinking of it as a "pain", and learn to embrace it. Trust me, you'll be glad you did.



So you know, ppi is absolutely meaningless in this context. All that matters is the actual number of pixels you're working with, not how many of them happen to fit into an inch. Ppi, or dpi, only matters for print. It doesn't matter in the slightest for on-screen work.

Even the number 72 is totally arbitrary in today's world. Many years ago, it was a loosely employed standard, intended for Web designers to get a sense of how large an image would appear on most users' screens. But modern displays are highly variable, there is no single number of pixels per inch that can be applied to anything. From cell phones to smartphones to laptops to desktop monitors to HDTV's, people are using all kinds of devices to view digital imagery these days, and none of them have any singularly binding standards.

But even forgetting about all that for the moment, the fact remains that at no time in history did ppi ever bear any relevance at all for texture work. The 3D models textures get applied to neither know nor care what an inch is. All they care about is how many pixels you're putting on them. It's all about pixels per surface, nothing more, and surfaces can be literally any size, or no size at all, depending on the context.

Consider this. Put a 512x512 texture on a cube. If that cube happens to occupy one square inch on your screen, then the texture is currently 512 ppi, for all intents and purposes. (Whether your monitor can actually display 512 ppi is an entirely separate question, of course.) Now zoom in on the cube, so it occupies your entire screen. If you've got a 24" monitor, it's probably around 21 inches wide. You're now seeing that texture as 512 pixels per 21 inches, or about 24 pixels per inch. Who's to say what the actual ppi of the texture is in this context? Really, there IS no actual ppi. The term is completely meaningless in 3D.


As for down-sizing, well, yes, you're going to lose information from your image whenever you do that (again, regardless of ppi). If downsample a 1024x1024 to 512x512, you've removed 75% of the pixels. By definition, there's no way to do that and preserve every last detail. Removing pixels is removing pixels is removing pixels.

The key is to design your imagery from the start with downsampling in mind. If you know your target size is going to be 1/4 of your source size, then you know that nothing smaller than about 2px in width is going to survive the downsizing process. So make your details accordingly. If there's something you know you'll want to preserve, make sure it's at least 4 pixels square. Anything smaller will either become fuzzy and faded, or will simply disappear entirely.

One of the primary advantages of working at a large source size, as I'm sure you know, is it greatly reduces margin for error. This is especially true with clothing textures. If you're trying to line up a seam at 512x512, and you're off by a pixel or two, you're going to see that on the model. But if you're working at, say, 2048, and you're off by that same amount, there's almost no way the difference could survive the downsizing process. Once the image has been resized to 512, that seam will be in perfect alignment.

But large source imagery can never allow you to cram any more detail into the smaller target image than you otherwise could. If the target is 512 pixels wide, it's 512 pixels wide, and that's that. You have to begin with the end in mind, always, or you will encounter problems.

i never said i didnt understand aplhas i said they are a pain in the ass.

Chosen, you misunderstood my boggle and then lashed out in many directions.
pay attention. First of all stop talking to people like they are stoopid.
i wasnt asking to be taught, i was asking to compare notes, get thoughts, OPEN MIND. i was using alphas and targas when kiki was the wipe on the newtek video toaster.

My concern was in the downsampling from a larg working file ie 1024 x 1024 @ 300 ppi or even 150. Thats all. some people like to work in 512 and even you mention some that like to work in some rediculously large number. i was just wondering how others dealt with the loss of pixels combining. if instead of working with a 3 pixel wide line for example knowing that the pixels will be combined with any compression... do has anyone designed with 4, 3 or 2 pixel lines at 300 dpi to get a clean 1 pixel line when it was downsampled to 75 ppi.

you dont have to reply to me Chosen.. im sure you have better things to do.
Peace

I would submit that if you truly understood them, you wouldn't find them to be a pain at all. If you're having any difficulty with them whatsoever, then you're undoubtedly missing something. There's simply no other way to explain it.

That's not a slight against you. It's a simple assessment of how these things work, based on many years of teaching this material to countless thousands of people. The workflow is super easy. If someone thinks it's at all hard, then there's something about they don't yet fully understand. That's all there is to it.




I'm sorry if that's how you chose to interpret it, but there was no lashing, outward or otherwise, from my direction. I was simply explaining how things work. You can decide to view that as an attack on you if you really feel the need, but that certainly wasn't what was actually happening.




So let me get this straight. If I've misinterpreted you, it must be because I "wasn't paying attention", but when you completely misinterpret me, as you did, it's because I'm "lashing out"? Doesn't that seem just a little bit one-sided to you?

Of course I was paying attention. That's a given. Do you really think it's even physically possible for someone to read through a post without paying attention to it? If you can do that, then I'd have to say it's a pretty neat trick, because I certainly can't.

Look, If the message you tried to send wasn't the one received, then there are only two possible explanations. Either you didn't explain what you meant clearly enough for me to follow, or you did but I somehow managed to misunderstand it anyway. Either way, lack of attention has absolutely nothing to do with it.

Please don't come into stickied threads with such confrontational comments. There's simply no place for that here.




You really think that was my intention? Come on, man. Look, if you come here and ask a question, I'm going to answer. And my answer will ALWAYS be addressed to the general public, not just to the specific person who asked it. Sometimes that might mean I'll explain a few things you already know. That's OK. It doesn't mean you're being called stupid. It just means I'm trying to give the most complete answer I possibly can, so that EVERYONE reading, not just you, can get the most they can out of it.

Remember, discussions here are for the entire population to read, forever, not just for the few who may be actively engaged in the discussion at any given time. If that offends your sensibilities, then I'd recommend you either adjust your own thinking, or else seek help elsewhere than public forums. Either way, please don't try to derail stickied threads by making the discussion about you instead of about the topic at hand.




I really don't see the difference. Comparing notes IS teaching. If we compare what we both know and don't know, chances are pretty good we're each going to have at least a little knowledge the other doesn't. If you tell me something I didn't know before, you've taught me, and if I tell you something you didn't know before, I've taught you. That's just how communication works. Recognizing that is a large part of what it means to have an open mind in the first place.




If that's the case, then I can't for the life of me imagine why you'd think they're so difficult to use. If I really had to guess, I'd suppose it's probably that it was poorly explained to you way back in the beginning, and as a result, it's been much more challenging for you than it ever should have been, all along. But whatever the problem is, provided you yourself actually do have the very open mind you just implied you value so highly, then I'm sure we can solve it, and you'll be able to use alphas with just as much ease as the rest of us.

Bottom line, if you think there's anything even remotely painful about it, then there's something you've missed in how it works. It really couldn't be easier.

Now, once again, you can choose to view that as an assumption that you must be "stoopid", as you put it, or you can see it for what it really is, a sincere offer to help you. The choice is yours.




As I said earlier, ppi is in no way relevant in this context. I'm not sure why you seem to keep insisting that it is. Again, all that matters is the total number of pixels you're working with. How many happen to fit into an inch only matters if the image is to be printed. Beyond that, your computer neither knows nor cares what an inch is.

Also, let's try and keep our terminology straight. Downsampling and compression are two different things. Yet you appear to be using the words synonymously. As long as you do that sort of thing, I've got no way of knowing what you're actually trying to say.




No, I don't HAVE to reply to you, or to anyone else, for that matter. But I do. I volunteer my time here every day, whether you think I should have "better things to do" or not. My reasons aren't your concern. You can either accept the information I freely offer, or you can leave it. Either way, please don't introduce such confrontation into stickied threads. The material here is too important to too many people for it to be diluted in that way.

your right. i didnt explain myself right the first time. let me start again


is there anyone that is working in photoshop cs4 with starting files (ie clothing template) that are 1024 x 1024 @ 150 dpi.? Then save them to png? Do you reduce the working file in the conversion to 72 dpi or have you been leaving them as 150's.

unless you are doing this or have any experience doing this please do NOT answer me. as i want nothing to do with alphas.


Peace
Zoro

Thank you for the apology, and for clarifying your question.

Now, for the third (and hopefully final) time, let me state with no uncertainty that dpi (or ppi) has absolutely nothing to do with texturing. Your image could be 1 dpi or 100 dpi or 1,000,000 dpi, and the texture will be exactly the same. Again, the ONLY number that is relevant is the actual amount of pixels you're working with. How many of them might happen to fit into an inch only matters if and when the image is being printed. For on-screen work, it's absolutely 100% completely meaningless.

Ask any texture artist in the world that same question, and you'll get the same answer. You're simply barking up the wrong tree by continuing to fret over dpi. It's irrelevant.

That said, if you simply must have a direct answer to your direct question, then there are actually three answers, not just one. They are yes, no, and it doesn't matter in the slightest. Yes, if you want to change the dpi from 150 to 72, you can; go ahead. Your texture will behave exactly as you want it to. And no, if you don't want to change the dpi setting, then don't. Your texture will still behave exactly as you want it to. Either way, it won't make any difference whatsoever. Dpi is simply a piece of metadata that instructs printers what to do. It has absolutely no bearing at all on on-screen imagery, textures included. 3D engines aren't even aware of its existence at all.

What WILL make a difference is if you change the number of pixels that make up the image. If you rescale from 1024x1024 to 512x512, for example, that's significant. You will have cut the amount of resources the texture consumes by 75%. This will be equally true no matter what dpi setting you'd care to choose.


That is the absolute truth of the situation. If you have any further questions on this particular subject, I'm going to ask that you post them in a new thread. There is no point in continuing to post the same thing over and over and over again in this sticky.

i will start a new thread as you are soooo wrong about ppi's relevance.

Looking forward to the thread. I can't imagine how DPI could possibly be significant for SL, where it's all about the pixels.

I'm looking forward to the new thread as well. Not to toot my own horn, but I've been doing this (for a living) for quite a few years now, and I cannot fathom how dpi or ppi could be relevant to texturing in any way, so I'm very curious to see your full explanation of how you think it does relate.

If it turns out there's any actual significance, we should by all means come back here and add the relevant information to the sticky. But understand, you'd be the first person in the history of graphics to find such significance, which would be quite an accomplishment.