what would couse this?

i was making placecards for my vendors. i did all the work for them in 1024 by 1024 and then resized them to 256 by 256 then uploaded and they look fuzzy i'm not taking just the picture part which i could see being fuzzy if you went from bitmap to tga but the lettering as well i did everything in photoshop then saved to a 24 bit tga file and uploaded the pictures all started as a 1024 by 1024 as well so no resize up then down any guesses

A lil secret that I use for making TGA files is this. In photoshop create your 512x512 72dpi template. Then resample it up to 300dpi, which is standard print resolution and gives you a clearer image and text to begin with. Do all your work at 300dpi, cut your alpha map at 300 as well, this is absolutely crucial. Flatten it down and save all your work. At the last step downsample it back to 72dpi, or screen resolution and save an alternate copy to upload into SL. Cutting alpha channels at print resolutions gives a lot of ghosting. I have even down sampled them smaller, to 256 x 256 for particles and TGA branches and leaves for trees, with absolutely no apparent ghosting or blurriness. The more times you manipulate an image in photoshop the more chance you get for distortion so always do your work at a higher size then save an alternate copy in an appropriate size at the last step.

that basically what i did was worked bigger than saved it as a smaller size it's still at 1024 on my computer i just uploaded it at 256

also would taking the origianl picture of the item that i'm making the sign for at say 2000 by 2000 and then at the last point drop it to 256 using an even number like that would help finding true center easier is why i ask
does it have to be on an 8 bit number cycle or can i use a diffent bigger number to start with and lower it ot 256 or 512

Wow, there's a ton of misinformation in this thread so far. Let me clear up the erroneous info first, and then I'll talk about possible causes of your problem, Mjolnir.



Huh? Since when does changing formats from BMP to TGA make things any more or less fuzzy? TGA is just a simple bitmap format, just like BMP is. You could switch an image back and forth between the two a thousand times and it won't affect the way the thing looks at all.

Did you perhaps mean something else?

Changing to JPEG will make things fuzzier, as JPEG uses lossy compression, but TGA should never do that. If that's what normally happens for you, I'd say maybe you've got a graphics driver problem. That's the only explanation I can think of.


No offense, Atom, but almost everything you said in your post is incorrect. I'll take it one point at a time.



Completely untrue. Dpi makes absolutely zero difference for texturing. Dpi only matters for print, not for anything on-screen. You could work at 300dpi or 72 or 1000 or 1 or .01, and it will make absolutely no difference. All that matters for texturing is the total number of pixels in the image, not how many of them fit into an inch. Textures have no idea what inches even are. All they know about are pixels.



I wanted to extract this part about resampling to make something clear. If you change the dpi without resampling, it won't change the number of pixels in the image, and so will have zero effect on texturing. However, if you do up-sample from 72 to 300 as Atom's post instructs (again, no offense intended, Atom), you'll be increasing the number of pixels by more than 16 times. That will make everything significantly blurrier. Those extra pixels need to come from somewhere, so what happens is the original ones get spaced out, and then the empty space in between gets filled in with interpolated colors. There's simply no way to do that and retain full clarity. It cannot be done.

So, all you accomplish by this up-sampling is to make your image less sharp, not more. If you want to start with a large number of pixels right off the bat, and then down-sample later, that would be OK, but never up-sample.



NEVER EVER EVER flatten!!! ALWAYS preserve your working document as a layered PSD. TGA's are inherently layerless, so flattening accomplishes nothing. As I so often say whenever this subject comes up, your source PSD could have one layer or one million layers, and your TGA would never know the difference. The TGA will come out exactly the same, no matter how many layers are in the working document.

The ONLY thing you do achieve by flattening is to make it harder to make changes later if you need to. There's nothing worse, for example, than flattening a sign, only discover later that you've got a typo you didn't notice before. Your only recourse at that point is to make the whole damned thing over again. Had you preserved your layered work though, you could just correct the spelling in 2 clicks, re-output a new TGA, and be done.

There was a tutorial that popped up a few years ago, in the early days of these forums, instructing people to flatten before outputting to TGA, and ever since then, some people just believe blindly that this step is necessary. I've been doing my best to combat that false belief ever since. It's been mostly stamped out, I think, but it still pops up every so often (like now).

Don't flatten.


Normally, down-sampling is relatively trouble free (within reason, of course), but doing it after having previously up-sampled can be quite problematic. When you up-sampled, you added all kinds of interpolated data to the image, blurring it. Now, when you down-sample, you're removing pixels, and there's no guarantee the ones that remain will be the original ones you started with. Even if they were though, all you'd accomplish would be to remove the very same interpolations you added before, making your whole process nothing but a giant waste of time. But what's more likely is that you'll end up with a different set of pixels from the actual ones you started with, meaning you've now got a "copy of a copy", which will look quite noticeably worse than the original.

Don't do all this crazy resizing, and as I've said several times now, never be concerned with dpi for on-screen images. Again, all that matters is the number of pixels. Inches have nothing to do with it.

On a side note, even though 72 dpi is considered "standard" for screen resolution, it's just a made up number. It was never actually representative of how many pixels really exist in a single inch of computer screen. Modern flat panels tend to have anywhere from about 65 to 130 pixels in each inch. CRT's have an even broader range, as they're capable of operating at multiple resolutions. The number 72 is pretty much completely meaningless.


What do you mean by "ghosting"? Do you mean haloing?

And what do you mean by "cutting" alpha channels? Do you mean making them?

If you do indeed mean to say that making an alpha channel in a 300 dpi image will produce more haloing than in lower dpi images, that's simply not true. The only way it could kind of sort of maybe be interpreted that way is if you use the logic that having more pixels in an image means having more of EVERYTHING, including the pixels that make up a halo. However, the percentage of halo pixels to all the other pixels will always be the same.

In any case, preventing haloing is a simple process, that has been written about at length, time and again, all over this forum.


256x256 or smaller should be the size of most of your textures. Going bigger than that is almost always completely unnecessary. You seem to be implying that 256x256 is very small. I would encourage you to read my sticky on texture sizes at the top of the forum. 256x256 is actually the most common size for game textures, and absolutely should be the most common size for almost everything you make for SL. For little things like particles (most particles are little, anyway, and don't require much detail), you can go MUCH smaller.

Never use more pixels in a texture than you actually need. The biggest reason SL runs as slowly as it does is that too many people use textures that are far larger than they should be. Read the sticky for more on this.



This is true. That's why I find it so puzzling that you would have adopted this bizarre up-sampling/down-sampling work flow of yours. Either work at the desired output resolution from start to finish (usually the best thing to do) or else work at a higher one and then only down-sample once at the time of final output.



If you're gonna do the latter, you'll get cleanest results by working at even power-of-two multiples of your desired output size. For example, if you intend your output to be 512x512, like for clothing, you might want your working document to be 1024x1024 or 2048x2048 or 4096x4096. That way, when you down-sample, you'll have even division of pixel groupings, which will yield the best chance for cleanest results.

I must repeat though, dpi has no place in that equation. It's all about the number of pixels, not about physical units of measure.




OK, now that all that's out of the way, let's talk about your problem, Mjolnir. You're saying your images look fuzzy after upload, not before, right? If so, that just confirms what I said earlier about TGA. If the problem were with your TGA file itself, you'd see the blurriness right in Photoshop. Since, by your description, that's not what's happening, you can rest assured that your TGA's are behaving properly.

As for why an image might look blurred in SL, there are many reasons. First, SL uses JPEG2000 compression to store images. (Your actual TGA file never gets uploaded. It gets copied to JPEG2000 as soon as you hit the Upload button.) While JPEG2000 compression can be lossless, SL does not use the lossless option for most images. (Images 64x64 and smaller are supposed to be lossless, so that sculpt maps will import cleanly, but that functionality is currently broken, so everything is lossy.) Generally speaking though, the loss is slight enough that most people would never notice it. It shouldn't be very obvious.

Second, when you upload an image, SL creates several different versions at various levels of detail (this is called "progression". Notice how when you go to a new area for the first time, textures on objects in that area first look blurry, then get a little clearer, and then a little clearer than that, and then finally reach full clarity? That's because it's cycling through the progressive LOD image sequence. This prevents you from having to stare at a blank screen while the full image is being delivered to you. You get the low-res version first while the next higher version is loading.

Usually all versions of the image are created within a minute or so of upload, but at times of heavy traffic, the final version can take up to a day or so before it's generated. So if a texture is giving you trouble today, take another look at it tomorrow before you decide it's actually bad.

Third, when apply a texture to an object in-wrold, you're almost guaranteed to be seeing it at a size other than its actual resolution. For example, if you put a 256x256 on a cube, and zoom in on it, that 256x256 is being stretched to fill the whole screen. If your SL window is 1024 pixels wide, you're looking at your texture at roughly 16 times its actual size. Generally SL is exceedingly good at displaying this kind of thing VERY clearly, but you can't expect it to be razor sharp.

Fourth, anisotropic filtering is your friend (as long as your system can handle it). Without anisotropic filtering turned on, textures on objects that are not directly facing the camera will always look blurry. AF corrects for this, keeping everything sharp, no matter what the angle. Go into your graphics preferences and turn AF on if it's not on already (again, assuming your system can handle it).

Aside from those four things, which are just SL's natural behavior, there's no reason your uploaded textures should look noticeably blurrier than your local image files. If you don't think any of those things describe your problem, post some screenshots to show us first hand. I'm sure we can pinpoint the problem if we can see it.

ok so the best way is what i was doing basically saved the bitmap at 1024^2 worked it at that level then reduced it to 256^2

now i did notice that the text looked better when i riased the base size to 500 and use a small pt as opposed to using it at 125 and using 72 pt even if it was close to the same size on the art work i don't know why this wokred it just did

so can i save the photos from sl at like 1000 work them that size then reduce to 256 or 512 without issue becouse 1000 is easier to find center than say 1024?

I'm not sure what you mean by "base size", so I'm not really following you in that second paragraph.

As for your other question, yes, you could start at 1000 instead of 1024, but the downsampling to 256 won't be quite as good from 1000 as from 1024. Practically speaking it's close enough probably that no one would notice, but in principle, it's not the best thing to do. I'm kind of a stickler for principles.

I'm a little but perplexed as to why you feel that "finding the center" of 1000 is easier than "finding the center" of 1024. Simply set your rulers to pixel mode, park a couple of guides at 512 on both axes, and you're all set. And/or just turn on Smart Guides, and you'll be given visual cues as you drag things around the canvas to show you what lines up with what. And/or turn on visibility for the grid, set the grid size to any power of two (I use 128 usually, with subdivisions set to , and you can snap things to halves, quarters, or eighths of the canvas very, very quickly and easily.

Just break yourself out of that base-10 prejudice that was ironed into you in grade school, and open your mind more to base-2. It's an important habit to develop.

well it's beingable to see where the line is in middle a 5 is easyer to see than a 3 the 3 lines are much smaller but i will try setting it up like you say that appears to be logical

ty

Maybe I wasn't quite clear earlier. You should never really have to worry about the ledger lines on the ruler. There are lots more ways to go than just the few I mentioned. Here are some more.

Turn on Snap To Rulers, and then simply drag a guide (or any other object with a smart guide) and. As soon as you hit the center line, the object will snap into place. So don't worry

Or, if you prefer not to drag the guides into place, you can manually assign them their locations with text. Just go View -> New Guide, and then plug the numbers into the dialog. Create a horizontal and a vertical guide, both at 512, and where they meet is your center point.

Or, you can track your exact cursor position by looking at the Info pane. Click on the Info tab at the top right, in palette with the Navigator and Histogram. Move your cursor around the canvas, and see what happens next to the X and the Y on the Info pane. Dee those numbers? That's the position of the cursor. Move the cursor to 512,512 and you know it's in the center.

You can also just pop up a little indicator to show you the exact center of your canvas at any time. Select the entire canvas (Select -> All, or ctrl-A) and then hit Free Transform (Edit -> Free Transform, or ctrl-T). The transformation handles will appear, including the centerpoint marker.

There are lots and lots of ways to find the center. I think simple math is the easiest and most logical, but do whatever you like best.

a trick i use to help things out is to oversharpen

sharpen your image to just below the point where edges start to lighten up, i personally do this in the gimp, so in that software i usually go for about 30-35% added sharpness, you will have to feel around for your desired effect with your software

it will look abit ugly inside your editor, but were not going to save this part to the source PSD file, do it as a last step before you save the file your going to send to SL

the general idea here is between open GL, and JPEG 2k, the image is going to fuzz out a little bit, and this is just a way to help compensate for it

alternatively the same effect can be done using various edge filters, such as in paint shop pro, edge enhance filters, often time reducing the steps down to 1 click vs. many in that program

heres an example, this test tile is a 128x128 24 bit bmp, expanded to roughly 4x its real size in SL, the one on the left was uploaded without extra sharpening, the one on the right had 35% sharpness added to what it already was before upload (using the gimp)

Work in the actual size of your document, resizing down causes textures to blur. If you want the end texture to be 256x256 then work in 256x256. If you work in 1024x1024 and then shrink down to 256x256 you are trying to squeeze 8 pixel worth of varied color into 1. If you work in 512x512 and squeeze down to 256x256 you are trying to cram 4 varied color pixels into 1.

oooooooooooooooo ..... You can see that the block on the left is 16x16 pixels.
oooooooooooooooo ..... This is called a power of 2 texture, which you get when
oooooooooooooooo ..... you take 2 and multiply it by 2 ( 4 ) and then continue to
oooooooooooooooo ..... multiply that result by 2. 2, 4, 8, 16, 32, 64, 128, 256,
oooooooooooooooo ..... 512, 1024, 2048, 4096, 8192, etc.. why do games work
oooooooooooooooo ..... in textures that are power of 2? Has to do with how the
oooooooooooooooo ..... computers handle storage of data and too long to go into
oooooooooooooooo ..... here. What you need to know involves why your texture
oooooooooooooooo ..... is getting blurry when you shrink it. Here's why. If you
oooooooooooooooo ..... Look at the *example* on the left you can see 16x16 of
oooooooooooooooo ..... our make believe pixels. Now lets imagine that each of
oooooooooooooooo ..... the *pixels* are a different color ( I can't show that on
oooooooooooooooo ..... these forums but I imagine you can work with the idea ).
oooooooooooooooo ..... When you shrink ^2 texture in half ( 512 to 256 )you are
oooooooooooooooo ..... in essence cutting it by 1/4 it's size. Thus trying to stuff
oooooooooooooooo ..... four different colors into one. This is what causes the
.......................................image to blur and lose it's sharp definition.

....................................The above 16^2 would be the 512^2 and the small one
oooooooo ...................below ( 8x8 ) would be our 256^2 that you are turning it
oooooooo ...................into. So what would it be like going from 1024^2 to a
oooooooo ...................256^2 you ask?
oooooooo
oooooooo ...................o o o o This is the 256 now and the 16x16 is the 1024^2
oooooooo ...................o o o o so you can see that you are REALLY cramming
oooooooo ...................o o o o that detail into the smaller space ( 8 pixels of
oooooooo ...................o o o o color into 1 colored pixel ).

So if you would like some advice from someone who does texturing for games for a living, work in the correct size so that what you see is what you get.

I just wanted to elaborate on this a bit since it's something a lot of people don't understand. DPI is simply an instruction used by printers and has no bearing on anything else. It tells the printer "print this many pixels per inch on the page." So if you have an image that's 3000 pixels wide with a DPI setting of 300, a printer will print an image 10 inches wide with 300 pixels per inch. If you changed it to 600 DPI (without resampling) the printer would print an image 5 inches wide with 600 pixels per inch, yielding a sharper but smaller image. When doing texture work that's not going to be printed and is only intended for dispaly on monitors the DPI setting is meaningless and you can completely ignore it. As Chosen said, the only thing that matters is the number of pixels in the image.

If I may continue the discussion on resolution specs and how they affect image editing...

First of all, thanks to Chosen Few for another classroom-in-a-post he made in this thread.

OK, we know that dpi settings have to do with printing images so let's leave dpi out of the discussion since my questions below have to do only with image manipulation that will be seen on a computer screen.

Computer monitors show a certain number of pixels per inch depending on their manufacturing specs. As Chosen Few mentioned above, the number of ppi that a monitor can show can range from 60 to 150 for modern flat panels depending on their size, native resolution, etc. So, that 72 ppi setting that is so popular in image creation mostly for the web, etc, is an arbitrary setting that ended up being somewhat of a standard. Fine up to here.

The confusing thing for me is when I create a 100x100 image at 72 ppi in my image editor and then another 100x100 image at 150 ppi they both look exactly the same on my screen and everyone else's too - they're both 100x100 images.

When I save them in their native file format (Paint Shop Pro) they are both exactly the same size, meaning they have the same number of pixels, which I assume to be 1,000.

So how does that initial 72ppi and 150ppi change the image? If the defining factor in the final number of pixels is the 100x100 initial size, what does the ppi setting do to the image and what effect does it have on how it's seen on computer monitors?

When I know that I'll know everything

it does nothing at all on computer monitors, 1 pixel in a image = 1 pixel on your screen

dpi is for print media only

Resolution is the most misunderstood term by people new to computer graphics, partly because it is an overloaded term (different people use the term to describe image dimensions, print dpi, screen dpi, size of the screen, etc.) I'm going to be teaching an intro Illustrator course (vector graphics of course make the whole discussion even more complicated) to graphic design students next semester and I'm figuring I'll need to devote an entire class session to resolution, plus explaining it again over and over throughout the semester. I'm sure the students will be bored out of their skulls by all the technical mumbo jumbo and I'm not really looking forward to that.

PCs got too popular too quickly and the graphics and printing industry didn't have much time to catch up and develop a standard way to refer to the specifics of their business in a straightforward and uncomplicated way (they're not the only ones of course).

Professionals in the field are as much confused by these terms as are people new to computer graphics. They may have get used to clicking on certain buttons in their application to produce the desired result but I've come in professional contact with graphics and designer professionals who are as confused by these things as most casual users of graphics editing programs.

The increased amount of business in the printing industry due to computer resolution mistakes from professionals is certainly not negligible. I've seen the tens of thousands of $ in extra monthly income first hand.

Yes, but image editing programs don't refer to dpi in their File > New dialog windows.

They refer to pixels per inch, or ppi as I called it, and that's what I was asking about.

If those two terms refer to the same thing, and I assume they don't, then that would be a definite and unavoidable source of confusion.

Thank you Chosen, Chip, Osgeld and maybe some others for explaining that dpi was merely an instruction to a printer. I've always been puzzled by what it meant.

This has been a very interesting and informative thread.

they are the same, in programs they use whatever they wish, most follow the photoshop standard of pixels per inch

eventho both are correct, they technically do 2 different things



but again were not printing

Wow - thanks Chosen and Chip - brilliant posts ! Can this be made a sticky?

Chosen - one VERY minor point..... when you said "However, if you do up-sample from 72 to 300 as Atom's post instructs (again, no offense intended, Atom), you'll be increasing the number of pixels by more than 4 times."....... I think that should be 16 times......

Oh right, even "print resolution" means two different things depending on if you are talking about the resolution of the image or the resolution of the printer.

!@#$!#@$%#$@^@#$%~!

this is really freakin simple

image resolution X pixels * Y pixels, nothing else 1 pixel = 1 pixel on your screen

ppi/dpi how many pixels per inch WHEN YOU PRINT IT, if your not printing it dont even think about it

the only other thing that DPI could mean is how densely your printer can print nothing else

So the pixels per inch number (ppi) that is in the File > New dialog windows of image editing programs is irrelevant to the image *until* the image is printed?

Is the purpose of that number only to tell the software how many dots per inch (dpi) to instruct the printer to print when the user prints the image?

If that's the case, and I have no reason to doubt it is, then the confusion continues and is fueled primarily by software developers who knowingly don't use consistent and specific terminology even in their own products.

If ppi and dpi is the same thing in an image editor then Adobe, Corel and everyone else should pick one or the other and stop being the source of the confusion.

And if that number has no relevance to the image until the image is printed out, why have it in the dialog window in the first place? What purpose is there to ask the user to specify a pixels per inch if that number comes into play only if the user prints the image? Ask for that number only when the user prints the image, not when they're creating it.

Most apps are designed with print in mind which is why the default rulers in PS show inches instead of pixels. The first thing I do whenever I do a fresh install of Photoshop is change the default ruler settings to show pixels instead.

The only way in which you need to factor in DPI settings when designing an image is if it's going to be printed at some point. You need to know the max dimensions and DPI it will ever be printed at so you can make sure you create it at sufficient resolution to avoid ever having to up sample. For example, if you do a full page ad for someone that's going to be printed 8.5x11 at 300 DPI you need an image that's 2550x3300 pixels (plus whatever bleed area the printer requires), and you want to be reasonably sure they aren't going to turn around and ask you to do a 40x60 inch poster from it later. Sampling down is fine and won't degrade your image much. Sampling up is bad mojo. There are plugins like Genuine Fractals that do a considerably better job of up sampling an image than the default methods in Photoshop, but it's still something you want to avoid if you can. All of this applies mainly to raster images. Vector graphics programs are what's mainly used for print design work because they're resolution independent.

Good point.

In the case where someone creates graphics for viewing only on a computer screen:

Would it be a good idea then to create all images at 300 pixels per inch and forget that arbitrary 72 pixels per inch that about 99% of all tutorials on image editing for the web instruct to use?

For example, a 100x100 image has 10,000 pixels whether or not I set it at 72 PPI or 300 PPI when I create it. Correct?

If that PPI number comes into play only when I print the image, then why not create everything at 300 PPI, to better match an expected average printer setting of 300 DPI that's in most home printers, just to avoid the case of having to print out an occasional image and not having to up-sample from one that was created at 72 PPI?

If the PPI setting is not affecting the image in terms of computer resources then even a 600 PPI setting across the board for all images intended for a computer screen would be the same as if they were created at 72 PPI. They wouldn't require more computer memory or storage space because their no of pixels would remain the same at any PPI setting. Is that right?