tga v png

I have noticed in the last few tutorials by Torley that he advocates uploading textures in PNG format.I have always used tga ,does anyone know the pros and cons?

He does? *facepalm*

Just the other day I was discussing the issue with Lazure in the last few posts of the SLCP thread, and Chosen discusses this at length in the alpha transparency thread.

I just noticed he doesn't mention PNG in the first post though, only later on in the thread, so maybe he should update that first post... Regardless, most of what he says about Photoshop 7 in that first post applies to PNG; in essence, what Photoshop 7 did was make TGA behave like PNG.

Thanks i will check that out

Or you could just try searching the forum archives for "TGA," "PNG," "alpha," or other relevant key words. This topic seems to come up several times a month.

However in my experience (and I could be doing something wrong) Even after setting my image to alpha, if I save it in .png, it does not upload to sl in alpha form. If I change it to .tga, I'm alpha'd all the way. I use PSP8.

I do not believe there are alpha channels in .png. You have to use simple transparency by changing opacity in the image. This is the advantage of png in SL's context, for simple transparency.

Yeah, I've been meaning to talk to him about that.

It's not that PNG is necessarily bad, of course. It's just that the way Torley talks about TGA and alpha channels in his recent videos is pretty irresponsible. Not only is his information inaccurate, it borders on actively discouraging. He describes alpha channels with such phrases as "old school", and "too complicated", if I remember correctly. Not cool.



You're probably right that I should update that. I've thought about it a few times, but I haven't actually done it (obviously). When I wrote that post, SL did not yet support PNG. TGA was the only option for transparency.

I'd like to update the information in both the FAQ post and the tutorials post. I know more about PSP now than I did then, so I could do some better writing for it, and I think all tutorials could benefit from an overhaul. Some of procedures could be presented in better ways, I think.

It's just a question of finding the time to redo it all.

Chosen

Call me lazy but I wanted to read what you had previously added in your other posts re PNG without having to search.

Pehaps you could add a link in this thread for other readers.

In the meantime I'll go search for them.

Larrie, there may or may not be a discussion about it on the alpha channel thread. I don't remember off hand, and I don't feel like looking for it myself right now. Press ctrl-F in your browser, type PNG into the Search field, and quickly scan through each page. You'll find whatever anyone has written about it, including me, right away. That's the same way I'd find whatever's there if I were to look.

There is some information on it in the file formats thread.

There is an old tutorial thread in here in which it was discussed. It wasn't in the alpha channel thread. The tutorial had someone who made stained glass in a png.

I read all the relevant stuff stuff as directed and despite Torleys viids have to decided to keep with tga. I have no problems with it at all.I just thought this was perhaps a change of track by SL and that png was now the preferred choice.Obviously not.

I think it is wrong to debate about "alpha channels" vs. "simple transparency" in this way.

For one thing, BOTH TGA and PNG files use "simple transparency"...that is, they do not have a separately editable alpha channel, as you find in a PSD file.

Nontheless, they DO have alpha channels, just not editable ones. *ANY* image that supports transparency has an alpha channel. The alpha channel is simply an additional 8 bits per pixel of information that pertains to the pixel's transparency, as opposed to its red, blue, or green color values.

32 bit TGA files and PNG files are, as far as I can see, identical in their utility to us. SL accepts them both (though I've heard that it likes PNG a little better; maybe it is a closer match to SL's native JPEG 2000 format). Both are what I would call a "distribution" format. Only save your work to that format when you are done and ready to export it to SL. All actual work should be done in a program and a file format that supports layers, including alpha channel layers.

I'm not sure where you're getting that, Lindal. You can absolutely edit the alpha channel directly, in any image that has one, regardless of format.

If you're using Photoshop (any version other than the highly flawed 7.0, that is), just open the file, go to the Channels palette, and it's all right there. You can edit any or all channels at any time, including alpha channels, with all paint tools, filters, etc. Editing an alpha channel is no different from editing any other element of an image. The file format does not and cannot change that.

I'm not sure what abilities GIMP has in this regard, but I'd be quite surprised if you can't do the same thing in it. I know its channels palette implementation is a little screwy, but still, you should be able to edit something as simple as an alpha channel any way you want. The file format shouldn't make any difference to whether you can or can't.

PSP has no channels palette, so you have to edit by proxy (copy channels to and from masks), but even in that program, the alpha channel doesn't become uneditable just because the file is in a particular format. An image is in an image, and a channel is a channel.

Or am I misunderstanding what you're saying?


I think the people who say "SL likes PNG better" are those who like it better themselves, so they come up with mythical justifications in their heads, and convince themselves it's the truth. All things being equal, an SL texture sourced from a PNG won't look any different from one sourced from TGA.

The potential problem with using PNG, though, is that things are not always equal. There's a greater margin for error with it. Because it employs simple transparency, if so much as a single pixel is less than 100% opaque, you end up with a 32-bit texture. For this reason, accidental 32-bit textures are quite common with PNG.

That can't happen with TGA (once again, except in the uber screwup of all time that was version 7.0), since it's pretty hard to create an alpha channel accidentally. It's a deliberate process that you either do or don't do, by deliberate choice.

The other arguments against simple transparency (relative lack of control, longer time requirements for all but the very simplest of images, etc.) have been discussed at length time and time again in other threads, and the OP appears satisfied at having read some of them, so I won't repeat them again here.

That isn't really a PNG thing, it's the application software. Photoshop's Save As... for PNG wants to include an alpha channel automatically if it appears to be necessary. For PS and the other Adobe apps, Save for Web... lets the channel be explicitly turned on and off, and the surprises go away.

That was my experience with PSP8, as well. I never could get them to work, so I just kept on using TGA to create transparencies. But, as John and Chosen would no doubt point out, it was probably good for me

Wow, fantastic tip! Thanks, Viktoria. It never even occurred to me to try that.

The other arguments against simple transparency still stand, of course, but that particular headache can be checked off the list. It's a bit silly that they don't put that transparency checkbox on the regular save dialog.

I am very amateur with Photoshop but when SL started allowing PNG I gave it a try. Hated the way PNG acted. I was getting alpha sorting issues with a PNG with no alpha. Having read the previous post I think I know now what happend. I must have had a pixel or two that were being read as transparent - I use a lot of layer of semi transparent layers to sometimes make a color or shading effect. I went back to only using TGA because - like mentioned - I make the alpha channel and even get one last chance when saving - 24 bit or 32 bit so I know what I will get in SL everytime.

Just my little input - the only thing I noticed that was good about PNG - they seem to draw faster when using high rez versions

Updating would be cool! I am trying to learn how to do this in Elements...it gets rather confusing to swap back and forth between the different sections since you have to read about the bikini a number a paragraphs up and then swap down to the other sections.

I do have to say everyone that contributes really makes this helpful to those of use trying to learn various things and cuts down on some of the learning and frustration curve even when I have no clue what everyone is talking about.

Have a great one all!

Thanks, Chosen...I didn't know you could do that in Photoshop with a flat image. I'm just a Photoshop dabbler, so I bow to your vast experience.

I had this same debate with Chosen a year ago, and made the tutorial that Torley linked to in his Giant Leek tutorial as a result of those discussions.

It's a little disappointing to read that the sticky still hasn't been updated to include information on how to make transparent files with .png, a year on since they were introduced. I haven't read the full 19 pages, and if it's in there, great. It should be part of the first page.

Siddean, while I'm certainly open to including some information about PNG in the sticky, I think it's important to keep in mind that, as its title suggests, the thread is about alpha channels. Any information added about simple transparency would only be for the purpose of enhancing readers' understanding and appreciation for why alpha channels are so important, and to demonstrate in no uncertain terms how they are so much more useful, powerful, flexible, and easy to use, than simple transparency could ever be.

As for your "disappointment", I'd invite you to consider three things. First, you seem to be speaking as if you think the thread is somehow official, and as if updates to it are something to which you and other readers are somehow entitled. Look, I'm just a "private citizen" here, just like you. Everything I do on these forums I do out of my own sense of helpfulness. I'm not official in any capacity, and I have no responsibility to do anything at all, other than that which comes out of my internally driven sense of duty toward helping my fellow human beings. The information I choose to share is what it is, and whether it's been a year or ten years or five minutes since alternative methodologies happen to have been added to they system, it doesn't change that. I'll get to revision when I get to it, and that's that.

Second, it's not like the alpha channel information will ever be out of date. The concepts behind them, and the techniques for using them, have been the same for decades, and will never go away. To "update" the post wouldn't mean changing any of the actual information. It would simply mean refining its presentation, a second draft, if you will.

Third, even if the thread weren't specifically titled "Transparency & Alpha Channels", I would still seriously have to question whether the addition of dedicated tutorials on simple transparency would be necessary. The reason alpha mapping needs to be explained is because it's not immediately obvious. But simple transparency is absolutely obvious, hence its name. Nothing could be quicker to understand than "create your image, erase the background (or don't create a background in the first place), save as PNG". I really don't think anyone could ever not be able to figure that out on their own.

That's not to say in any way it wasn't generous of you to have written a tutorial on it, yourself, or that you shouldn't have done, so don't get me wrong. It's really great that you did. As a member of the community, let me sincerely thank you for having taken the time to do it. I only wish everyone were so giving.

I do have a small handful of critiques on your tutorial, though, if you don't mind. Overall, it is quite well presented, but there are a few parts to which I have to take issue:


1. Thumbnails

Rather than spending the opening section complaining about things like how Windows does not come with TGA thumbnail support out of the box, why not offer solutions? Greggman's thumbplugs have been freely available for years now, and have been linked on these forums many, many times. Here's that link one more time: http://www.greggman.com/pages/thumbplug_tga.htm

Another options is ThumbView, which not only adds thumbnail support for TGA, but for 18 other image formats as well. http://www.seriema.net/thumbview/

Note that both options work by adding the necessary functionality directly to the Windows shell. They are not programs, and so do not need to be running in order to work. That means they don't consume memory or CPU power, and they don't require any actions on the part of the user, other than simply to install them. They serve to patch the OS itself, which a one-time operation, and then the thumbnails just work.

Why Microsoft chooses not to include full thumbnail support with Windows in the first place is anyone's guess. But I would submit that just because MS makes bad decisions is no reason to compound the problem by suggesting to people that they shouldn't use the most common staples of the entire graphics industry. Fixes are readily available.



2. Halos

You're certainly right that there's no need to add dehaloing layers or anything like that with PNG. But what you seem to be missing (which is partly my fault) is that there's no need for it with TGA either. I'll explain.

The tutorials in the sticky do include dehaloing steps, but that's only because the order of operations presented in them lends itself to haloing. If you were to reverse a few things, the halos would never be created in the first place, so there would never be any need to take the extra step of removing them. I chose to use the order I did, though, because I feel it's the easiest approach for the novice reader to understand, conceptually. In actual practice, none of those tutorials represent the way those who really know what they're doing would work. Where I feel short was in not explaining that right from the start, and in not offering more advanced tutorials after the beginner level ones.

It happens every so often here on this forum that someone will say "Wait a minute. If I do this instead of that, I notice it's impossible for a halo to form. I would assume that means I don't need to do the dehaloing thing. Am I right?" My answer to that is always "Congratulations. You've now moved past the level of relying on the tutorials, and you fully understand how this stuff works. Yes, you're absolutely right."

When I do get around to revising the sticky, I will certainly add some haloless tutorials. The halo-inclusive ones will have to remain, though, since they will always be the easiest ones for beginners to grasp. The haloless approaches are markedly simpler to perform, but a little more complicated to understand.



3. Plants (and foreground/background extractions in general)

I really have to tell you that in your decided attempt to avoid using channels, you turned what would have been a really simple channel-masking operation into a very unnecessarily complex process. By comparison the method you've presented is quite time-consuming, and very, very labor-intensive. There's absolutely no need for a simple extraction to be that complicated.

If you're interested, here's how to do the job much more quickly and efficiently, via a channel mask:


1. Take a look at each of the image's three color channels individually. Pick which one that has the most contrast, and duplicate it.

In the case of your wreath image, the channel with the most contrast is the Blue channel. Here's what that it looks like on its own.



2. Open the Levels dialog (Image -> Adjustments -> Levels), and use the Input sliders to increase the contrast as much as possible, without losing detail around the edges of the subject.

For your wreath, raising the black slider to about 170 or so does the trick. Here's what the channel looks like now. Notice it's already starting to look like a proper alpha channel.



3. Since we happen to be working with a dark subject against a light background, the we've got our blacks and whites opposite of how we'll want them in our finished alpha channel. Let's correct that now by inverting the colors (ctrl-I).

That's more like it.



4. Now just run over the edges of the subject with the Burn tool, at a high exposure setting. This will remove things like the existing shadow, which is gray in the channel, as well as those pesky in-between-the-pine-needles areas. This takes about two seconds, and you'll find that it completely negates the need to zoom in and erase anything. This is where the time-savings can be absolutely huge.

You'll also want to use the Dodge tool, and/or a white paintbrush to remove the stray black dots in the subject. This too will only take a couple of seconds.




5. At this point, you're done. You've got your alpha channel, exactly the way it needs to be. Save as 32-bit TGA, and you're all set.

I've taken the extra step here of copying the channel to layer mask, just to show how the transparency looks.



It might be a little hard to tell from just these pictures, but because we dodged the edges well in the alpha channel, there's no halo. It is possible, however, that due to JPEG compression artifacting, if you were to download the alpha channel image I posted here, and apply it, you could end up with some small degree of haloing here and there. Be aware that that's only because I happened to have posted the images here in lossy JPEG format, for the Web. There is no haloing whatsoever in the original.

Optionally, if you want to be doubly certain there's no halo, it's easy enough to slap a green background underneath the main layer, or you can apply the alpha channel as a mask, apply the mask to delete the background, and run the Soidify filter. Either one would add all of 2-3 seconds to the work time.

Total elapsed time for this, not counting posting the screenshots to the Web, and writing all these words, was about one minute. There was no need for extra layers, for zooming in and out, or for erasing anything. For this type of extraction, channel masking is fast, efficient, and accurate. There's no better method.

A couple of questions. Where is this filter in PhotoShop CS? and Can you merge two alpha channels together?

It's not in Photoshop unless you add it. It's a Flaming Pear filter. Check Robin Wood's web site = http://www.robinwood.com/Catalog/Technical/SL-Tuts/SLTutSet.html.

I'm not so sure I agree with your reasoning here. Bear in mind that I've not read the tutorials in question so I'm just speaking from conjecture based on your description, but still I don't think it is necessarily wise to start people out with a method that could result in halos. Even if you point out that it's a purposely roundabout beginner method and the REAL method doesn't need any dehaloing steps (and if you don't point this out, aieee!) people will get the wrong idea about things.

As you point out, special work to remove the halo with TGA is unnecessary, so why not just teach it right in the first place? It's not like it's markedly more complex (indeed, it's simpler) and I think if explained properly it's not too bad of a conceptual leap. Like, in the last couple posts of the SLCP thread I described alpha channels with the metaphor of clothing patterns.

Incidentally, I've seen the Flaming Pear plugin mentioned before and always kinda felt like that's overkill. I mean, you also mentioned the much simpler (in the sense that you don't have to install a new plugin) approach of simply creating a green background layer; I would suggest that is the approach people should be suggesting to newbies, since the plugin just makes things seem more complicated than they really are.



Sure. Alpha channels are just like any other image channels, and so can be painted or blended or whatever just like any other part of the image. Now I forget if you need to copy the alpha channel to a layer to do your work and then copy it back to the alpha channel, or if you can blend channels together directly, but it's certainly possible and quite easy.



Never mind the plugin, this is a great tutorial in general for anyone confused about alpha channels. I very much agree with her assessment that the main reason people think alpha channels are hard to work with is that they don't understand what a "channel" is, and her tutorial does a great job of de-mystifying that basic computer graphics concept.