Second Life Forums Archive - what would couse this?

Chip Midnight

ate my baby!

Join date: 1 May 2003

Posts: 10,231

12-01-2007 16:37

From: Daz Karas

n 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?

The best thing you can do is just forget about it completely. If you absolutely have to put something in there, use your favorite number, like maybe you're mom's birthday or something. The only thing you need to care about is the raw pixel dimensions of the image. You can always change the DPI setting later if you decide to print the image.

_____________________

My other hobby:
www.live365.com/stations/chip_midnight

Daz Karas

Registered User

Join date: 14 Feb 2007

Posts: 53

12-01-2007 17:14

From: Chip Midnight

...The only thing you need to care about is the raw pixel dimensions of the image. You can always change the DPI setting later if you decide to print the image.

Now we're getting somewhere. Thanks for the tip.

Chosen Few

Alpha Channel Slave

Join date: 16 Jan 2004

Posts: 7,496

12-01-2007 17:21

From: Rudolph Ormsby

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......

Ooh, good catch, Rudolph. Yes, you're right, of course. Thanks. I'll edit the post.

From: Daz Karas

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?

As Chip said, you can safely just forget about it completely. However, I still use 72 for on-screen images just as a way of labeling them as for the screen. When people encounter a 72 dpi image, they assume it's not meant for print.

While saving at 300 wouldn't make any difference to the way it appears on-screen, it could make a difference to how people perceive its purpose. I'd just hate to hand someone a 300 dpi image, have them mistakenly expect to be able to print it because it's got that label, and then they discover it comes out the size of a postage stamp on paper because it was only 256x256 or something. That's all.

I think of the 72 as almost a metadata tag, saying "don't print this".

From: Artemis Tairov

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.

Technically, you're correct, Artemis, and for many types of images, I agree; work at your intended output size from start to finish. However, there are many situations in texturing when it can be highly advantageous to work big and then downsize for output.

SL clothing is great example. Matching lines and other distinctive imagery across seams is a particular challenge, which is very easy to screw up. On a 512x512 canvas (the native size of all clothing textures), with the way the canvas is chopped up for the various body parts, you only have a very small number of pixels in each area. In many places, being off by as much as a single pixel is enough to make the seams not match. That's a huge margin for error.

When you work on an upsized canvas though, the margin for error gets cut by 75% for every doubling of the canvas size. So if you're working at, say, 2048x2048, and you're off by one pixel on a seam, it won't be noticeable after you downsize to 512. The very data blending you're talking about as bad actually becomes a good thing in cases like this.

The slight downside of this is you just need to be careful not to make your details too small. Obviously, if you're working at anything larger than the final output size, and you include details that are at all smaller than the size of the pixel groupings that are going to be combined by the down-sampling, then those details won't be present in the final image. So be a little careful. Don't put a super fine cloth weave pattern on a 2048x2048 clothing template, and then wonder why it looks like just a flat field of solid color after it's downsized to 512.

Generally, the trade off better margin for error so you can do it right vs. potential detail loss if you happen to do it wrong is well worth it. You just gotta know what you're doing (which is why we have these discussions

).

_____________________

.

Land now available for rent in Indigo. Low rates. Quiet, low-lag mainland sim with good neighbors. IM me in-world if you're interested.

Mjolnir Uriza

Hammer of the Gods

Join date: 14 Sep 2007

Posts: 504

12-02-2007 06:12

ok thank you all i have been trying several of the sujestion with great results i still use the 1024 at 72 and reduce and it is very clear when i use black lettering on a colored background still can't get white lettering to look right but that might be a fact that the colors are to far apart and the system is trying to cominsate for something

ok one last make sure i understand ok. tga all the way and don't use jpg right becouse jpg loses resalution

Johan Durant

Registered User

Join date: 7 Aug 2006

Posts: 1,657

12-02-2007 06:58

"loses resolution" isn't really what is happening with jpegs, and you should be saving to psd while working and only something else in the end for final output, but yes output tga for images to upload into sl.

file formats checklist:
psd -- use while working on the image
tga -- use for uploading to sl
jpg -- do not use at all for working with sl

_____________________

(Aelin 184,194,22)

The Motion Merchant - an animation store specializing in two-person interactions

Rachel Darling

Registered User

Join date: 3 Jun 2006

Posts: 95

12-05-2007 08:09

From: Osgeld Barmy

a trick i use to help things out is to oversharpen

I believe this is good advice to follow. I frequently use close-up photos of fabrics and trims, which then have to be downsized to get the proper scale for SL clothing. Reducing the size of the element by 50% or more tends to "fuzz" the image and lose detail, so typically I do a Sharpen, a Sharpen Edges, or an Unsharp Mask operation (depending upon which one looks best for the element). Oversharpening beforehand often gives me the best results. Finally, when I reduce the final clothing template from 1024x1024 to 512x512 just before creating the alpha prior to exporting to tga, I do a final Sharpen on any highly-detailed element layers.

Insofar as jaggy text, I always do my text in as large a size as possible, and refrain from rasterizing the text layer. If you rasterize the text and then resize, you'll typically get more jaggies -- especially with smaller or narrow fonts. I also try to adjust the font size to whole points after resizing (rather than 36.7, for instance), though I'm not sure this really makes a huge difference with the current font formats.

Infrared Wind

Gridologist

Join date: 7 Jan 2007

Posts: 662

11-03-2008 06:30

Methinks one must be patient for the server to fully load the texture.

I just did some tests. Uploaded a JPG version. It was blurry.

Then uploaded a PNG. Still blurry.

Then a TGA. It was crystal clear.

Then another TGA (all the same file, different names.) And it was blurry.

More on this tomorrow!

- I

_____________________

VonGklugelstein Alter

Bedah Profeshinal Tekstur

Join date: 22 Dec 2007

Posts: 808

11-03-2008 07:42

From: Mjolnir Uriza

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

My 2 cents:

The subject went off track .. so I am reading this as .... The text is fuzzy on my 256 x 256 cards for my vendors

True! while 256 x 256 will work adequately for mickey mouse looking walls and building textures - you can't really expect a bunch of text to be legible on such a small area unless you are only doing a few words using larger font.

1 Always try to use high contrast colors to separate your background from your font color.. ( I like to use white font on Black Background so I can then tint it and make the text any color I want... if I want

2 Use a font that is as simple and square as possible... something like Arial works better than some scribbly doo fancy girly font .. Keep it simple is the best approach for info cards

3. If you have to use a larger texture to get the info to be clear.. DO IT!!! use a 512 or even 1024 if its super detailed. If the product in your store is nice stuff, people will chill and wait for the texture to load.

4. Preview your texture before upload - do you have any idea how many millions of uploads are mistakes? If it looks like crap at say 300-500% on your screen - it will most likely look like crap on SL.. my oppinion only of course . but it seems to work - I hardly get any surprises with regards to overall quality of detail that way

All that other technical stuff is just fluff... I hear the texture size argument all the time because I tend to keep my textures on the larger size although I have been offering "LOW RESOLUTION" versions of some of the textures to offer the choice for those who do not know how to compose balance between how many textures to use in a build and creative use of a great texture - and then blame the big textures for lag.. (haha)

I have been using a tiling viewer to preview some textures in question before uploading.. when you tile something you can get a good sense for the balance and quality of an image.
Its a great way to see if you need the detail or not.. and size accordingly

\m/

_____________________

FD Spark

Prim & Texture Doodler

Join date: 30 Oct 2006

Posts: 4,697

11-05-2008 23:26

It seem like when I try to save layer without something goes wrong on my end
I have TGA files that refuse to upload unless merged.
I have had hidden layers show up on the transparent uploads.
I have had transparent layers go totally transparent.
When I have tried to do tiny city a while back ago I had these tiny window
views that I spent literally hours trying to do elaborate
windows views inside the building.
The windows inside I figured they need to be at least 32 pixels.
I divided 512 space up by how many windows, bricks,etc.
I thought got all areas measured out by pixels what I was going to draw in it.
I am not sure if I resized when it was tga or not but it all the window views
became all blurry once uploaded.
And then I got really cranky about doodling any more tiny city textures and I gave up for bit. Or least until I started clothes now I am running into this issue with if I want very tiny but detailed items.
Some one suggested certain tools when reducing. It some where on buttons.
I have to look it up later.
Any I always thought tinier the image less the resollution.
So is this isn't true?
I have friend who does very beautiful flexible prim dresses she says jpgs are best
tga cause problems with flexible skirts and won't use tga or png on skirts
she does. There are times I guess some people use jpg's here.

_____________________

Look for my alt Dagon Xanith on Youtube.com

Newest video is

Loneliness by Duo Zikr DX's Alts & SL Art Death of Avatar

what would couse this?
Chip Midnight ate my baby! Join date: 1 May 2003 Posts: 10,231	12-01-2007 16:37 From: Daz Karas n 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? The best thing you can do is just forget about it completely. If you absolutely have to put something in there, use your favorite number, like maybe you're mom's birthday or something. The only thing you need to care about is the raw pixel dimensions of the image. You can always change the DPI setting later if you decide to print the image. _____________________ My other hobby: www.live365.com/stations/chip_midnight
Daz Karas Registered User Join date: 14 Feb 2007 Posts: 53	12-01-2007 17:14 From: Chip Midnight ...The only thing you need to care about is the raw pixel dimensions of the image. You can always change the DPI setting later if you decide to print the image. Now we're getting somewhere. Thanks for the tip.
Chosen Few Alpha Channel Slave Join date: 16 Jan 2004 Posts: 7,496	12-01-2007 17:21 From: Rudolph Ormsby 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...... Ooh, good catch, Rudolph. Yes, you're right, of course. Thanks. I'll edit the post. From: Daz Karas 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? As Chip said, you can safely just forget about it completely. However, I still use 72 for on-screen images just as a way of labeling them as for the screen. When people encounter a 72 dpi image, they assume it's not meant for print. While saving at 300 wouldn't make any difference to the way it appears on-screen, it could make a difference to how people perceive its purpose. I'd just hate to hand someone a 300 dpi image, have them mistakenly expect to be able to print it because it's got that label, and then they discover it comes out the size of a postage stamp on paper because it was only 256x256 or something. That's all. I think of the 72 as almost a metadata tag, saying "don't print this". From: Artemis Tairov 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. Technically, you're correct, Artemis, and for many types of images, I agree; work at your intended output size from start to finish. However, there are many situations in texturing when it can be highly advantageous to work big and then downsize for output. SL clothing is great example. Matching lines and other distinctive imagery across seams is a particular challenge, which is very easy to screw up. On a 512x512 canvas (the native size of all clothing textures), with the way the canvas is chopped up for the various body parts, you only have a very small number of pixels in each area. In many places, being off by as much as a single pixel is enough to make the seams not match. That's a huge margin for error. When you work on an upsized canvas though, the margin for error gets cut by 75% for every doubling of the canvas size. So if you're working at, say, 2048x2048, and you're off by one pixel on a seam, it won't be noticeable after you downsize to 512. The very data blending you're talking about as bad actually becomes a good thing in cases like this. The slight downside of this is you just need to be careful not to make your details too small. Obviously, if you're working at anything larger than the final output size, and you include details that are at all smaller than the size of the pixel groupings that are going to be combined by the down-sampling, then those details won't be present in the final image. So be a little careful. Don't put a super fine cloth weave pattern on a 2048x2048 clothing template, and then wonder why it looks like just a flat field of solid color after it's downsized to 512. Generally, the trade off better margin for error so you can do it right vs. potential detail loss if you happen to do it wrong is well worth it. You just gotta know what you're doing (which is why we have these discussions ). _____________________ . Land now available for rent in Indigo. Low rates. Quiet, low-lag mainland sim with good neighbors. IM me in-world if you're interested.
Mjolnir Uriza Hammer of the Gods Join date: 14 Sep 2007 Posts: 504	12-02-2007 06:12 ok thank you all i have been trying several of the sujestion with great results i still use the 1024 at 72 and reduce and it is very clear when i use black lettering on a colored background still can't get white lettering to look right but that might be a fact that the colors are to far apart and the system is trying to cominsate for something ok one last make sure i understand ok. tga all the way and don't use jpg right becouse jpg loses resalution
Johan Durant Registered User Join date: 7 Aug 2006 Posts: 1,657	12-02-2007 06:58 "loses resolution" isn't really what is happening with jpegs, and you should be saving to psd while working and only something else in the end for final output, but yes output tga for images to upload into sl. file formats checklist: psd -- use while working on the image tga -- use for uploading to sl jpg -- do not use at all for working with sl _____________________ (Aelin 184,194,22) The Motion Merchant - an animation store specializing in two-person interactions
Rachel Darling Registered User Join date: 3 Jun 2006 Posts: 95	12-05-2007 08:09 From: Osgeld Barmy a trick i use to help things out is to oversharpen I believe this is good advice to follow. I frequently use close-up photos of fabrics and trims, which then have to be downsized to get the proper scale for SL clothing. Reducing the size of the element by 50% or more tends to "fuzz" the image and lose detail, so typically I do a Sharpen, a Sharpen Edges, or an Unsharp Mask operation (depending upon which one looks best for the element). Oversharpening beforehand often gives me the best results. Finally, when I reduce the final clothing template from 1024x1024 to 512x512 just before creating the alpha prior to exporting to tga, I do a final Sharpen on any highly-detailed element layers. Insofar as jaggy text, I always do my text in as large a size as possible, and refrain from rasterizing the text layer. If you rasterize the text and then resize, you'll typically get more jaggies -- especially with smaller or narrow fonts. I also try to adjust the font size to whole points after resizing (rather than 36.7, for instance), though I'm not sure this really makes a huge difference with the current font formats.
Infrared Wind Gridologist Join date: 7 Jan 2007 Posts: 662	11-03-2008 06:30 Methinks one must be patient for the server to fully load the texture. I just did some tests. Uploaded a JPG version. It was blurry. Then uploaded a PNG. Still blurry. Then a TGA. It was crystal clear. Then another TGA (all the same file, different names.) And it was blurry. More on this tomorrow! - I _____________________
VonGklugelstein Alter Bedah Profeshinal Tekstur Join date: 22 Dec 2007 Posts: 808	11-03-2008 07:42 From: Mjolnir Uriza 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 My 2 cents: The subject went off track .. so I am reading this as .... The text is fuzzy on my 256 x 256 cards for my vendors True! while 256 x 256 will work adequately for mickey mouse looking walls and building textures - you can't really expect a bunch of text to be legible on such a small area unless you are only doing a few words using larger font. 1 Always try to use high contrast colors to separate your background from your font color.. ( I like to use white font on Black Background so I can then tint it and make the text any color I want... if I want 2 Use a font that is as simple and square as possible... something like Arial works better than some scribbly doo fancy girly font .. Keep it simple is the best approach for info cards 3. If you have to use a larger texture to get the info to be clear.. DO IT!!! use a 512 or even 1024 if its super detailed. If the product in your store is nice stuff, people will chill and wait for the texture to load. 4. Preview your texture before upload - do you have any idea how many millions of uploads are mistakes? If it looks like crap at say 300-500% on your screen - it will most likely look like crap on SL.. my oppinion only of course . but it seems to work - I hardly get any surprises with regards to overall quality of detail that way All that other technical stuff is just fluff... I hear the texture size argument all the time because I tend to keep my textures on the larger size although I have been offering "LOW RESOLUTION" versions of some of the textures to offer the choice for those who do not know how to compose balance between how many textures to use in a build and creative use of a great texture - and then blame the big textures for lag.. (haha) I have been using a tiling viewer to preview some textures in question before uploading.. when you tile something you can get a good sense for the balance and quality of an image. Its a great way to see if you need the detail or not.. and size accordingly \m/ _____________________
FD Spark Prim & Texture Doodler Join date: 30 Oct 2006 Posts: 4,697	11-05-2008 23:26 It seem like when I try to save layer without something goes wrong on my end I have TGA files that refuse to upload unless merged. I have had hidden layers show up on the transparent uploads. I have had transparent layers go totally transparent. When I have tried to do tiny city a while back ago I had these tiny window views that I spent literally hours trying to do elaborate windows views inside the building. The windows inside I figured they need to be at least 32 pixels. I divided 512 space up by how many windows, bricks,etc. I thought got all areas measured out by pixels what I was going to draw in it. I am not sure if I resized when it was tga or not but it all the window views became all blurry once uploaded. And then I got really cranky about doodling any more tiny city textures and I gave up for bit. Or least until I started clothes now I am running into this issue with if I want very tiny but detailed items. Some one suggested certain tools when reducing. It some where on buttons. I have to look it up later. Any I always thought tinier the image less the resollution. So is this isn't true? I have friend who does very beautiful flexible prim dresses she says jpgs are best tga cause problems with flexible skirts and won't use tga or png on skirts she does. There are times I guess some people use jpg's here. _____________________ Look for my alt Dagon Xanith on Youtube.com Newest video is Loneliness by Duo Zikr DX's Alts & SL Art Death of Avatar

Welcome to the Second Life Forums Archive

what would couse this?