Remove Rough Edges

[DQ Darwin]

Hi

I have looked, read, played, jumped up and down but I still get rough edges on my clothing. I have included an exaggerated pic to show the problem. This is a blouse neckline. I use Photoshop CS4.

Any hints or help or tips appreciated. Thanks

[valerie Inshan]

Hi Dee! The liquify tool in CS4 is my best friend (and rough edges' ennemy!)

[JubJub Forder]

For a start,  try raising the resolution of your file, its good to work at twice the final resolution and size down when finished.

Drawing smoothly by hand is real hard to do, so theres a variety of ways to get around it. My advice is learn to use the  Paths tools (in Photoshop)  for basic shapes. Or use software like Illustrator (vector shape specialist software)

 

[valerie Inshan]

I agree. Path tools are really an easy and efficient way to solve the problem, especially if tou slightly feather the edges (1,2 or 2 pixels, no more). And gaussian blur too works in many cases.

[DQ Darwin]

@Val: Hi Val and thanks I will give your tips a tryout. In fact the path tool is what I use but feathering the edges I didn't think of trying,

@JubJub: Hello, using Illustrator is an interesting option, thanks.

I thank you both for your tips and now I am off to play and see what happens.

.

[robin Meridoc]

Hi!

Early in my 'clothes making career' I had this problem too - and some kind soul told me about Gaussian Blur.  I use GIMP so I dont know what its called in PS but I'm sure there is an equivalent.  You'll have to experiment and find how many pixels you have to blur to make those jaggies go away.

If your fabric has some sort of pattern or texture to it you'll need to be careful with the blur.  What I would suggest in that case is to create a 'trim' that you can blur separate from the fabric.  You'll have to make your fabric slightly smaller then to lay in the middle of your trim.... ifyou dont the jaggies may show underneath the trim.

This is one of the two basic things you need to make clothes in SL.  The other is bezier curves, in GIMP they're called Paths.  I dont know what they're called in PS - perhaps someone can add a post with the proper term.   

I'll try to put a post on my blog on this in the next couple of weeks - using both Paths and Gaussian Blur in GIMP to make clothes.  http://robin-makingstuff.blogspot.com/

robin

[DQ Darwin]

Hi Robin: Thanks all the tips are great. By the way I have read your blog and want to say thanks here for all the work you put into it.

[Neelah Xue]

Don't use filters, firstly, despite what was told to you about "liquefy". That's a bunch of crap, and all that will do is blur edges, and still keep pixels visible. Put on a white skin and wear your clothing if you want to see what I mean. Just increase the resolution of your image when you paint it. I.E. paint it at 3000 x 3000 when you are using your template. SL will transition everything down to the right size while retaining a much higher detail edge.

[Josh Susanto]

I use Lunapic, not PS, but I agree with those who say you should size up you images before editing them.

Sizing them down later shouldn't really be necessary, but if you feel like you're wasting precious data space otherwise, then, sure. I usually size things up to the nearest exponent of 2 on each side and just leave them there. If it is something where I need a softer curve like in the given example, sizing it up as far as possible seems to be the basic idea. 

Mostly, I deal with natural rock surfaces. If they are 3000 on a side, I will tend to crop them to 2048 or some other number with a lot of factors of 2, and then shrink them to 1024 before loading, just so I can see what's going to happen to them if I load them. Always crop keeping in mind what SL is going to do. Cropping and resizing are not the same, and you can protect at least some of the image clarity by cropping to dimensions that SL is actually going to use, rather than dimensions it will fudge to fit.

One advantage to me of editing with an elarged image is that if I use the cutout circle function, a larger circle is a rounder circle. This matters because I produce seamless textures by offsetting images so that the edge becomes a cross in the center, and then dropping a corroded circle over the cross. I corrode the circles by applying a %-controlled magic wand successively to a series of color borders. Turning one pixel into 4 or more pixels before putting the circle crop through assures a more complex starting circle edge and gives me slightly more control of the final corroded contour.

There are also types of edit data that will do more than simply duplicate pixels if you increase the dimensions of the image before applying them. I'm not a big fan of PS liquefy, but I expect you'll do less harm with it if you first make the image as large as possible (think about it). The negative side effects of things like a "sharpen" tool can also be somewhat mitigated simply using an enlarged version of the initial image. Don't just believe me, though. Try it for yourself and compare.

[Chosen Few]

 

---

Neelah Xue wrote:

Don't use filters

---

 

I simply can't imagine why anyone would suggest that fuilters should not be used.  They're incredibly useful, for all kinds of things.  Filters are a tremendously important, arguably vital, part of the modern image creation process. 

If you haven't been using them, you've been missing out.  If you need help understanding what filters do, what they're for, how they work, and/or how to use them for specific texturing challenges, ask away.  There are plenty of us here who would be happy to walk you through it.

 

 

---

Neelah Xue wrote:

what was told to you about "liquefy". That's a bunch of crap,

---

 

It's not a "bunch of crap".  Just because you don't understand someone else's technnique doesn't mean it's invalid.  Obviously the person who posted the advice about the liquefy filter has been getting success from it, or she wouldn't have posted it.  Nobody's here to mislead anyone.

That said, I do agree with you that liquefy filter wouldn't be my first choice for dealing with alaisaing.  It's not that it wouldn't work; it's just that there are much more efficient ways to prevent the problem in the first place.

 

 

---

Neelah Xue wrote:

all that will do is blur edges, and still keep pixels visible.

---

 

That doesn't have to be true.  The liquefy filter is no different from any other tool, in the sense that it can be used literally hundreds of thousands of different ways, and for literally millions of different purposes.  You MIGHT end up with the results you describe, if you use it a certain way.  But you can also end up with all kinds of other results, by using it other ways.

My guess is that the person who suggested it has been using it to 'fold' the aliased edges into the filled areas, effectively removing the aliased pixels (jagged edges) from view.  There's no reason that wouldn't work.  Again, I'd rather prevent the problem in the first place, so that the whole liquefy step would be unnecessary, but that's personal preference.  If someone prefers doing it the harder way, then so be it.

Bottom line, there are a gazillion ways to skin a cat in Photoshop.  Don't EVER suggest that something can't work, just because you haven't personally experienced success with it, especially when someone else so obviously has.  Otherwise, your comments only serve to underscore your own lack of experience at best, and your own closed-mindedness at worst.

 

 

---

Neelah Xue wrote:

Put on a white skin and wear your clothing if you want to see what I mean.

---

 

I fail to see how just looking at results in isolation, whether good or bad, would explain "what you mean".  I coud use the liquefy filter right now to smooth out aliased edges, apply the clothing over a white skin, and completely invalidate everything you've said here.  Or I could botch the job, and seemingly confirm your claims.

All we can say with any certainty is that there are lots and lots of usable techniques.  For countless reasons, some tehcniques will jive with any given person, while others won't.  Please don't do anyone the disservice of insisting that your particular way is the only way.  We're here to discuss these things, and learn from each other, not bash each other for daring to use foreign techniques that we ourselves have yet to master.

 

 

---

Neelah Xue wrote:

Just increase the resolution of your image when you paint it. I.E. paint it at 3000 x 3000 when you are using your template.

---

Working at a large size, is almost always good policy, yes.  But I would NEVER go with an arbitrary number like 3000x3000.  For texturing, stick with powers of two, always.  The image is going to end up at 512x512 when baked into in-wolrld avatar outfit, no matter what.  By sticking with powers of two, from start to finish, you eliminate a huge of potential artifacting from uneven divisions in the down-sizing algorhithms. 

 

I usually work at 1024x1024, or 2048x2048.  There's not much point in going bigger than that, but if you really want to, the next step up would be 4096x4096.

In any case, I take issue with your use of the word "just" in this context.  You make it sound as if starting big is the only way to deal with the issue at hand.  You appear to be suggesting working around the problem, rather than tackling it head-on.

I can promise you, I can make a neck line at the native 512x512 that would look as good or better than any that could be made by "just" upsizing and then downsizing.  Again, my preference is to make sure the edges are anti-aliased properly in the first place, rather than realy on scaling to hide the problem after the fact.

There are a number of ways to do this (of course).  I usually work with the vector path tools, myself.  They're resolution-independent, so they provide a good means to create clean-looking lines at just about any size.  Further, they're procedureally transferrable from image to image.

And, of course, there's always good old fashioned hand-painting.  Use a mildly soft brush, and a Wacom tablet, and you can get super clean lines each and every time, with no fuss to speak of (assuming you're good at drawing/painting in the first place).

 

 

---

Neelah Xue wrote:

SL will transition everything down to the right size while retaining a much higher detail edge.

---

 

I don't ever want to surrender control of my imagery's appearance to SL, or to any other automated system.  I always want to directly control how lines are anti-aliased, myself.  I choose what kind of resizing algorhithms are in play.  SL doesn't get a say in that.  I make sure each and every pixel is how I want it, not how SL thinks I might want it, before I upload anything.

The purpose of working at a large size is not so that downsizing will clean up your edges.  They should be clean right from the start.  The reason to go big is so that you have more freedom of movement while you're working.  That's it.

You also get the side benefit of having more magin for error with things like seam matching.  If you're at normal scale, and a seam is off by one pixel, the mistake will be visible.  If you're at 4 times normal size (1024x1024), and you're off by one pixel, chances are very good that the mismatch will get averaged out when you downsize.

Now, don't get me wrong here.  I'm not advocating that such mistakes are OK, or that anyone should be cavelier about them.  You shoud always make every effort to make sure everything matches 100% perfectly, right from the start.  But the reality is we're all human, and such mistakes do happen.  Working at large scale increases tolerances.

 

 

 

Here's my general advice for how to create clean neck lines (and clean everyhting else):

1.  Work at 1024x1024, or 2048x2048.

2.  Draw the lines, exactly how you want them to be, with the pen tool. (And be sure to save the path, in case you want to use it again.)

3.  Fill the entire layer with the base color or base pattern you plan on using for the garment.

4.  Ctrl-click the path to form a selection from it, and then create layer mask from the selection.  The mask will "cut out" your edges.  (The reason I prefer masks is because they're non-destructive.  If you want to change the neck line later, you won't have to repaint anything on the layer itself.  Only the mask itself will need to be changed.)

You should now have a totally clean looking neck line, without having to downsize first, and without having go back in with the liquefy filter (or any other tool) to clean up your mistakes.  If it's not shaped quite how you want it, simply edit the mask accordingly.  You can do this by altering the vector path and repeating the process to make a mask from it, or simply by painting on the existing mask itself.

When everything is done, and looking exactly how you want it, then and only then should you downsize.  For best results, export to TGA at full size, and then open the TGA, and resize it to 512x512.  Keep your layered PSD at full size.  There's never any reason to shrink that.

[DQ Darwin]

Hi All. Again thanks for the tips. I have experimented with some already and see better results. The main ones so far have been size (now using 2048) and the path tool for clean cuts.

I will try liquefy, Gaussian blur and feathered edges among others.

Chosen you are totally correct and although some might I don't disregard any suggestion as each one is what works for the person posting and therefore has merit.

I like the idea of saving as targa bringing it back in and resize to 512x512 will try that also.

So far the difference is as below:

Done at 512

Done at 2048

[Josh Susanto]

Who would have imagined that just increasing the size of the image would actually cover more of the boob?

OK, I tease. But the edge does look quite a bit smoother to me.

I also wouldn't tell you not to try tools that people have reported as being effective.

If I have a clear reason for not using something, I will give that reason.

So far, does anyone have a reason NOT to resize things larger in order to get smoother curved edges?

[JubJub Forder]

Another hint:

Use .PNG instead of .TGA - you won't have trouble with white edges, managing alpha channel, etc.

[Josh Susanto]

Good to know.

The tools I have don't even work with TGA.

For file conversion, I use the free download from Converthub, which only does one file at a time, but is at least free.

A tip for anyone doing partially transparent PNG's is to convert them to TGA before you load them, or you'll tend to get black where you expect transparency checkerboard.

But you all know that already... right?

[Chosen Few]

 

---

Josh Susanto wrote:

So far, does anyone have a reason NOT to resize things larger in order to get smoother curved edges?

---

 

Good question.  I can give several reasons:

 

1.  It's completely unnecessary.  If you work properly, right from the start, to ensure that all your lines are anti-aliased as soon as you create them, then you'll never have to worry about this issue.  Resizing an entire image, just for the sake of smoothing a line that should have been smooth in the first place, is nothing more than a waste of time.

2.  It sacrifices control.  As I said in my earlier post, I want to directly control where each and every pixel in my images ends up.  I don't want to rely on automation for things like this, ever.  Uncertainty is simply not in my job description, and it shouldn't be in yours either.

3.  It can be very unnecessarily repetitive.  If after resizing, the line doesn't look quite how you might have wanted, you have to do the whole damned thing over again to make it right.  And if it's still not right, then it's rinse and repeat, and repeat, and repeat...  This can be tremendously time consuming, especially if you're working with a complex texture, or with a series of many textures.

4.  It's semi-unpredictable.  So far, we've been discussing this as if there will only be one line to smooth in the image.  Consider that a good game-quality texture is very likely to have dozens, if not hundreds, of individual layers in it, each of which will have elements that have their own individual edges.  Do you really want to have every line look jagged while you're working, so you have no idea what your final result is truly going to look like until after you've finished?  Not only would that be just a maddening, tear-your-hair-out, experience throughout the entire work process, what happens if you then, upon downsizing the image, discover that it looks like crap (which it likely will, if you've been working so blindly all along)?  Do you really want to have to start over again from scratch, only to repeat the same kinds of mistakes, all because you never bothered to learn how to prevent them?   With that approach, an image that should have taken a few hours to create could take days or even weeks.  Not cool.

5.  It's destructive.  As I mentioned earlier, I always want to work as non-destructively as possible.  I want infinite freedom to go back and change things as many times as I want, without making any sacrifices to quality and without any unnecessary addition of time.  Further, I always want to make sure that as many image elements as possible are reusable for other images.  For example, if I want two or more garments to have the same neck line, it would be a waste of time to have to create it more than once.  I'd rather just copy the line itself from one to the next.  The vector path/layer mask method I outlined is infinitely re-editable, and infinitely transferable from image to image.

6.  It can be overkill.  What if I've got an image in which I want smooth lines and jagged lines, both?  If I apply any process that smooths out everything, I'm screwed.  To borrow an expression from modern politics, we want to apply the scalpel here, not the machete.  Again, I want direct control over the smoothness and jaggedness of each of my lines, in real time.  The last thing I'd ever want to do is allow any automated process, be it a resizing algorithm or anything else, take away my powers of decision.

7.  Forgive the expression, but it's a "dummy's" way to work.  I can never sanction simply covering up a mistake with a band-aid. While it happens that it will often appear to be effective for the very simplistic kinds of imagery we've been discussing in this thread, that's really as far as it goes.  This kind of cover-up approach simply won't work for everything. if you don't learn how to create smooth lines from the get-go, you will fall flat on your face when you run into a circumstance in which resizing alone won't fix your mistakes.  I really hate to see people experience that kind of frustration.  Therefore, the only methods I will ever teach or recommend are those that are universally applicable.  The smart way to work, always, is to prevent a given problem from occurring in the first place, rather than just covering it up after the fact.

 

I could probably go on all day listing more reasons.  Here's something that might drive the point home, succinctly.  Try this.  Take a step back from everything you think you know about making textures, forget whatever habits you've picked up that are likely coloring your outlook about what is and isn't possible, and just think about the topic in the simplest possible terms for a moment.  Does it make any immediate logical sense at all that a line that is supposed to be smooth shouldn't just be smooth, right from the start?  I'm sure you'd agree, the only possible answer to that question is a resounding no, it doesn't make sense.  In simplest terms, if something is supposed to be smooth, it should just be smooth, period.

So, with this in mind, the only question worth asking becomes how do we make a line look smooth, right from the start.  The question of how to force a jagged line to become smooth really isn't even relevant.  If things are smooth in the first place, there's simply no need to even go there.

Prevent problems, and you'll never have to worry about covering them up.

 

 

 

---

Josh Susanto wrote:

I use Lunapic ...  The tools I have don't even work with TGA.

---

Josh, I can't stress this enough.  Do yourself a favor; get better tools!

Lunapic is a severely limited online photo editor, not a texture creation tool.  If you don't want to spend the money on Photoshop, that's understandable.  But GIMP and Paint.Net are both free, and Paintshop Pro is only $99.  All three of those options are full-featured image creation programs, more than suitable for high quality texturing at great speed.  Lunapic simply isn't.  The amount of time you've been costing yourself, and the limitations you've imposed on yourself, by using such an under-capable tool as Lunapic are staggaring to think about.  It almost makes me weep for you.

If the likes of Lunapic are what your texturng experience has been limited to thus far, then I have to say it makes sense why you'd be looking to options like resizing to solve your aliasing problems.  99.99% of the tools we'd normally talk about for preventing such problems (as well as for preventing tons upon tons of other problems) simply don't exist in applications like Lunapic.

Just so you know, I've been posting on this forum almost daily for the past seven years, and you're the first person I've ever seen say they've been using Lunapic for texturing.  I'm not even sure I've ever even seen it mentioned at all, come to think about it.  Unless you're a very dedicated massochist, get yourself something proper.

 

 

---

Josh Susanto wrote:

For file conversion, I use the free download from Converthub, which only does one file at a time, but is at least free.

---

Irfanview is also free, and it does batch conversions.  If a stand-alone converter is what you want, Irfanview is considered by many to be the best one out there.

Just about any full-featured image editor will, of course, also do batch conversion, including the aforementoioned GIMP and Paint.Net, which are both free.

 

 

 

---

Josh Susanto wrote:

A tip for anyone doing partially transparent PNG's is to convert them to TGA before you load them, or you'll tend to get black where you expect transparency checkerboard.

---

 

I've never seen that happen before.  It's more than likely a symptom of the tremendous limitations of the specific software you've been using. 

It sounds like the transparency in your PNG's isn't being generated in a way that the SL uploader fully understands.  The PNG format supports multiple forms of transparency.

[Chosen Few]

 

---

JubJub Forder wrote:

Another hint:

Use .PNG instead of .TGA - you won't have trouble with white edges, managing alpha channel, etc.

---

 

Oh, how my heart goes out to people refer to alpha channels and halo-prevention as "troublesome".  I've written so extensively on this topic, so many thousands of times over the years, on this very form and elsewhere, but still, it seems there will always be those who just assume that using alpha channels must be difficult, simply because it's not immediately obvious.  People end up making things so much harder on themselves this way, it's so sad.  All it takes is literally an hour or two of learning at most in the beginning to save countless thousands of hours over the course of your texturing activities.

The fact is it takes all of two clicks to create an alpha channel, once you know how.  And depending on how smart you work while you're creating the image itself, it's anywhere anywhere from zero to three clicks to make sure there's no haloing going on.

If you're using the WYSIWYG work flow, you're wasting so much time, and sacrificing so much control over what you're doing, it would make your head spin to witness even 1% of the benefits you've been missing out on. There's nothing faster, easier, or more reliable than alpha-map work flow.  That's why it has been THE staple of the entire graphics industry for the past 40 years, and why it will likely continue as such for countless decades to come.

And if we really want to do the logic behind it justice, it's worth mentioning that the very same kinds of processes were utilized in analog photography for centuries before the first computer was ever even dreamed of.  Trust me, if illiterate 18th century photographers in primitive homemade darkrooms could composite imagery by using masks, so can you.  There's really nothing to it.

If you need help, ask.  Don't just dismiss what you don't yet understand.

[DQ Darwin]

Well I for one find it very difficult to argue with logic so I won't. To me your point is clear and correct. Thank you Chosen:)

[JubJub Forder]

So lemme get this right - your recommending more clicks for a TGA file in one post, then in another saying don't resize cause it makes for more clicks if you change your mind?

We're trying to help a beginner with simple tips. 
PNG is a simpler format to use because it has less clicks for process. A format that you can easily see - no alpha channels hidden from view to trip a beginner.

You may recommend a beginner not avoid 'issues' and spend more time learning your 'proper way' before they get results - I don't.
I have 23  years experience with Pshop much of it fulltime and I still don't know every aspect of it - I don't pretend to... but i certainly don't try to help people learn by recommending harder ways, or things they don't actually need to know. It's hard enough to learn this stuff as it is.

A long time accepted, and general rule of thumb, is to start with a larger size and size down... you can argue the exceptions all ya want - its still a general rule of thumb thats useful for beginners and for most situations. And i point to the pictures above to prove our learner already has useful/better results.

Perhaps you've had to write this stuff  "thousands of times over the years" because you're over complicating things and people don't get it?

[valerie Inshan]

I just remember a great tutorial by Torley:

EVERY photographer should enable antialiasing at the source to smooth those jaggies! Instructions and a picture that shows a comparison are here.

Hope this helps! :smileywink:

[Chosen Few]

 

---

JubJub Forder wrote:

So lemme get this right - your recommending more clicks for a TGA file in one post, then in another saying don't resize cause it makes for more clicks if you change your mind?

---

You misunderstand.  Working with alpha channels isn't about "more clicks".  It's about avoiding all the hundreds of extra steps that are necessary throughout the entire image creation process if you're not using them.  It's about making things easier, not harder.

 

Say I want to make a simple image that grades from transparent to opaque.  If I've never been taught about alpha mapping, I'm going to have to spend hours with the eraser, subtly decreasing the opacity, line by line by painstaking line.  On the other hand, if someone was kind enough to explain the (exceedingly simple) logic behind how alpha mapping works, so that it has become an inherent, ingrained part of my basic thought process when approaching the entire subject if image creation, I'm just going to know instinctively that all I have to do is apply a simple black-white gradient to the alpha channel, and I'm done in less than a second.

You really want to tell me WYSIWYG is the easier method?  Again, there's a big difference between what's immediately obvious, and what's easy.

 

 

---

JubJub Forder wrote:

We're trying to help a beginner with simple tips. 

---

Exactly.  I've been helping beginners with this for more years and years and years. I can promise you from LOOOONG experience that teaching people good non-destructive work habits right from the start makes things infinitely easier on them.  By treating alpha channels as some mysterious thing that only advanced users would ever touch, you make things so much harder on the student, it's almost criminal.

I teach people about alpha mapping logic in their very first lesson, and from there, they never have to look back.  I treat it like what it is, the most fundamental part of how graphics works.  When the student isn't taught to see it as a big deal, they never become afraid of it, never think of it is hard or foreign.  They just use it, period.

Since simplicity is your goal, consider this.  It actually takes several times more chemical reactions in the brain to un-form a habit than to form one.  It's literally orders of magnitude harder for us to forget than to remember.  So, if we're presented with two options, ordered as "the beginner's way" and "the advanced way", most of us will just stick with the former forever, even if the latter is actually easier.  Once we've formed a habit, everything else becomes "hard", whether it actually is or not.

I take the subject of habit-forming very seriously in my teachings.  I don't want any budding texture artist ever to have to struggle to re-learn something that rightly should have been taught to them in the beginning.  There's simply no such thing as "beginners' methods" and "advanced methods". 

If you yourself were not taught about alpha channels right in the beginning, then it's possible that no amount of arguing from me will ever change your perception that they somehow have to be hard for newbies to understand.  But of that's indeed the case, then only underscores my point, as it means your own habits are firmly embedded, and can't be easily broken.

Here's something you might want to consider.  On the old, old SL forums (yes, there were old ones, and old old ones before them; these are the new ones), the question of how to create and work with alpha channels used come up literally several times a day.  It was the number one most frequently asked about subject, literally more common than all the other FAQ's combined.

So, I took it upon myself to write a "transparency guide", which explained what alphas are, what they're for, how they're used, and how to make them, in plain English, very newbie-friendly terms.  That guide became the very firs thread ever stickied in those old old content creation forums.  Within three days of its appearance at the top of the texturing forum, virtually all the questions about alphas stopped.  The frequency dropped from several times a day to once or twice every couple of months.

From then on, the answer was always just to ask, "Did you see the transparency guide at the top of the forum?" to which the reply would almost always be, "Oh, I didn't notice it.  Thanks for pointing it out," and then they'd get it.

To this day, I still get thank you notes from successful in-world designers (then newbies) who say they never would have learned to do the things they do had it not been for that guide.

I mention this not to toot my own horn, but simply to demonstrate just one example of how presenting people with the right information in the very beginning yields success.

When the old old forums were replaced with the old forums, stickies were no longer supported, so the guide disappeared into the archives.  Now that the new forums are here, and they do support stickies, perhaps it's time to resurrect it.  Clearly the need for it hasn't gone away.  It could use some updating, though, so I'll get on that as soon as I can.

 

 

---

JubJub Forder wrote:

PNG is a simpler format to use because it has

less

clicks for process. A format that you can easily see - no alpha channels hidden from view to trip a beginner.

---

I couldn't disagree more.  What if I'm a beginner, and I want to make something with varying levels of transparency, like a stained glass window.  It's seemingly simple in concept, but in actual practice, it's a somewhat challening item for a beginner to create successfully.

 

If I haven't been taught about alpha channels, it's going to take me hours upon hours to paint/erase each of those little panels with their graded opacities.  The image as a whole could take days to create.

Further, say I want to do three different versions of the same window, in three different color schemes.  If I'm working with WYSIWYG methodology, I'm going to have to repeat the entire thing three times.  I'm looking at potentially weeks of work, just for three little windows.  Man, texturing is hard!!!  Maybe I shouldn't even bother trying.

Now let's say a kindly instructor happens by, and informs me about how alpha mapping works, in terms that I as a newbie can easily understand.  Wow, now all of a sudden this project is easy.  It's only going to take me a few minutes to create all those different transparency levels, since I can just paint them into place, as shades of gray on the alpha channel.  Then I can copy that same alpha to each of the other images, and all I have to do to create the other versions is just change the colors.  The transparency levels are already taken care of.  Wow, this is exciting!  I can't wait to see what else I've been missing out on up until now!

I can't tell you how many times I've rescued people from utter despair by showing them how alphas can save the day so easily.  I make sure my own students never have to experience that kind of frustration, because as I said, they just don't know any different.  Alpha mapping is part of how they're taught to think about imagery, right off the bat.

I've said it a thousand times before, and I'll repeat it once again.  Just because something isn't necessarily immediately obvious doesn't mean it's not easy.  I promise you, WYSIWYG is almost always harder than alpha-map work flow.

 

As for your notion that alpha channels are somehow "hidden from view", I'm not sure where you're getting that.  In programs like Paintshop Pro, and PS Elements, maybe, since those programs lack a channels palette.  But in Photoshop, everything's right there.

For people using a sans-channels-palette program, I recommend putting all the layers in a group, applying a mask to the group, and using that as a proxy for the alpha channel.  When everything is done, simply save the mask as a channel.

This method works perfectly well in Photoshop as well, if anyone feels they need to see the transparency as they work, rather than just trusting the grayscale values to do their job.

So, nothing need be "hidden from view", ever.

 

 

---

JubJub Forder wrote:

You may recommend a beginner not avoid 'issues' and spend more time learning your 'proper way' before they get results - I don't.

---

Again, you misunderstand.  I don't recommend doing anything at all BEFORE getting results.  I recommend getting results right from the sart, by learning the most universally applicable ways to get them, in the very beginning.  I don't ever want to put someone in a position of forming a habit they'll later have to struggle to break.

 

 

 

---

JubJub Forder wrote:

I have 23  years experience with Pshop much of it fulltime and I still don't know every aspect of it - I don't pretend to... but i certainly don't try to help people learn by recommending harder ways, or things they don't actually need to know. It's hard enough to learn this stuff as it is.

---

 

It's great that you've accumulated so much Photoshop experience.  But how much of it has been in teaching budding texture artists?

From the way you talk about it, I would venture to guess that the majority of your experience in using photoshop has been in doing things like photo editing.  Photographers tend to be taught this stuff in a very different order from how texture artists are taught. 

I teach everyone the same way, including photographers.  But knowing how most photographers traditionally get presented with this stuff, I can appreciate why you're having hard time seeing what I'm advocating as anything other than bass ackwards.  I can assure you it's not.

Web designers tend to have it worst of all, in my experience, by the way.  They tend to be taught very linear methodologies, which aren't always the most practical when artistry is the goal. 

I had an interesting discussion with one just the other day, in fact, when he happened to stop by.  I was editing photos for one of my real estate clients (yes, I do more than just 3D work), and he was pretty dumbstruck by my methods.  The guy's been a photo editor for 20+ years, and he'd never been shown 99% of this stuff.  Like you, he thought alphas were for "advanced" people when he started.  End result, two decades later, he's still never embraced them.  When I think of all the over-billing to his clients, since he's racking up the hours by doing it the hard way with WYSIWYG, it's almost criminal.

 

As for teaching people things they don't need to know, there's no such thing.  It's ALL need-to-know.  That's not to say there's no order to how this stuff is best absorbed.  There certainly is.  But the notion that alphas can't come first is patently absurd.

 

 

---

JubJub Forder wrote:

A long time accepted, and general rule of thumb, is to start with a larger size and size down... you can argue the exceptions all ya want - its still a general rule of thumb thats useful for beginners and for

mos

t situations. And i point to the pictures above to prove our learner already has useful/better results.

---

 

Not sure where you're going with this paragraph.  I stated as much, myself.  OF COURSE working at a large size, and then downscaling afterward is almost always a good idea.  I never disputed that.  Quite the opposite, I encouraged it.

I just don't like when people present it as a crutch to work around problems that should never exist in the first place.

 

 

---

JubJub Forder wrote:

Perhaps you've had to write this stuff  "thousands of times over the years" because you're over complicating things and people don't get it?

---

 

No, I've had to write it thousands of times because there are always people out there who haven't read it yet.  That's just common sense. 

This subject appears and reappears all over the place, all the time, and I was just venting about that in what I thought was a an obviously tongue-in-cheek manner.  Perhaps the tone of the statement didn't come across as clearly as I'd thought it would. 

I certainly never meant to imply that it's the same people asking the same questions every time.  That would be pretty silly, wouldn't it?

I've never had anyone not get it.  I've had people refuse to even try to learn it, of course.  Unfortunately, there's no changing that in some people. But I've never had anyone actually sit down with the material, and go, "Uh, I can't understand this."  It's simple, and really easy to absorb, for anyone who wants to try.

Your stubborn insistence that this is "overcomplicated" is frankly insulting to the intelligence of every newbie out there.  You're presenting it as unnecessarily scary.  Your argument is troublingly pessimistic, and arguably detrimental to the open-minded learning process of anyone who might be seeking to learn this stuff for the first time.  There's absolutely no reason people can't learn to use alphas right from day one.  To insist otherwise, just because you yourself maybe didn't have that opportunity, helps no one.

[DQ Darwin]

@Val thank you for the link I will check it out.

@JubJub @Chosen: I would like to thank you both not only for the tips but the logic in discussion behind them. You both have given interesting points that I take well into account as I take further steps in my creative journey.

@Chosen: You are talking the path I agree with in so far as taking the time to get the fundamentals right and proceed from there. This implies more time at the start but hours saved in the future which is more to my way of thought. Though a bit foreword by me I do eagerly await your "transparency guide".

As said before I thank all for their respective input and appreciate the time taken to post here.

[Josh Susanto]

[1.  It's completely unnecessary.  If you work properly, right from the start, to ensure that all your lines are anti-aliased as soon as you create them, then you'll never have to worry about this issue.  Resizing an entire image, just for the sake of smoothing a line that should have been smooth in the first place, is nothing more than a waste of time.]

Which is why I specifically meant sizing it up BEFORE you do anything else. I guess if you're not working with templates or any other starting image, that would mean simply starting with a large canvas.

[2.  It sacrifices control.  As I said in my earlier post, I want to directly control where each and every pixel in my images ends up.  I don't want to rely on automation for things like this, ever.  Uncertainty is simply not in my job description, and it shouldn't be in yours either.]

I don't see how making 3 more of each pixel sacrifices control; it just splits each pixel, thus allowing 4 times as much control as before. I'm also not clear on how using a smaller image means less automation or less uncertainty unless we're talking about a non-duplicative resize that moves data across pixels.

[3.  It can be very unnecessarily repetitive.  If after resizing, the line doesn't look quite how you might have wanted, you have to do the whole damned thing over again to make it right.  And if it's still not right, then it's rinse and repeat, and repeat, and repeat...  This can be tremendously time consuming, especially if you're working with a complex texture, or with a series of many textures.]

I agree that it's unnecessarily repetitive if you do it AFTER you've tried with a smaller image and failed. That's a reason to enlarge FIRST.

[4.  It's semi-unpredictable.  So far, we've been discussing this as if there will only be one line to smooth in the image.  Consider that a good game-quality texture is very likely to have dozens, if not hundreds, of individual layers in it, each of which will have elements that have their own individual edges.  Do you really want to have every line look jagged while you're working, so you have no idea what your final result is truly going to look like until after you've finished?  Not only would that be just a maddening, tear-your-hair-out, experience throughout the entire work process, what happens if you then, upon downsizing the image, discover that it looks like crap (which it likely will, if you've been working so blindly all along)?  Do you really want to have to start over again from scratch, only to repeat the same kinds of mistakes, all because you never bothered to learn how to prevent them?   With that approach, an image that should have taken a few hours to create could take days or even weeks.  Not cool.]

The results of downsizing are unpredictable, true. But if you size something up from 512 to 1024 to edit and then leave it there, I don't see how that's a problem. Most of my upsizing is just to 512 or 1024.  I agree that upsizing a jagged line to smooth it rather than simply producing a smooth line in the first place is probably going to be the worst possible solution most of the time.

[5.  It's destructive.  As I mentioned earlier, I always want to work as non-destructively as possible.  I want infinite freedom to go back and change things as many times as I want, without making any sacrifices to quality and without any unnecessary addition of time.  Further, I always want to make sure that as many image elements as possible are reusable for other images.  For example, if I want two or more garments to have the same neck line, it would be a waste of time to have to create it more than once.  I'd rather just copy the line itself from one to the next.  The vector path/layer mask method I outlined is infinitely re-editable, and infinitely transferable from image to image.]

Merely duplicating a pixel is not destructive. Smearing it is destructive, but sizing the elements up once to the correct size assures that they only get smeared once per image, or not at all if the size-up is a simple multiple.

[6.  It can be overkill.  What if I've got an image in which I want smooth lines and jagged lines, both?  If I apply any process that smooths out everything, I'm screwed.  To borrow an expression from modern politics, we want to apply the scalpel here, not the machete.  Again, I want direct control over the smoothness and jaggedness of each of my lines, in real time.  The last thing I'd ever want to do is allow any automated process, be it a resizing algorithm or anything else, take away my powers of decision.]

It is overkill if it adds no utility to the specific edit process. Sizing everything up regardless of what you intend to do to it is pointless. Resizing is not smoothing. Smoothing is smoothing. It just works a little less sloppily if the image has twice as many pixels on a side. I generally dislike smoothing in any case and hardly ever see any reason to use it.

[7.  Forgive the expression, but it's a "dummy's" way to work.  I can never sanction simply covering up a mistake with a band-aid. While it happens that it will often appear to be effective for the very simplistic kinds of imagery we've been discussing in this thread, that's really as far as it goes.  This kind of cover-up approach simply won't work for everything. if you don't learn how to create smooth lines from the get-go, you will fall flat on your face when you run into a circumstance in which resizing alone won't fix your mistakes.  I really hate to see people experience that kind of frustration.  Therefore, the only methods I will ever teach or recommend are those that are universally applicable.  The smart way to work, always, is to prevent a given problem from occurring in the first place, rather than just covering it up after the fact.]

I do not recommend resizing up to cover mistakes. I recommend resizing up to prevent them in the first place.

 [Josh, I can't stress this enough.  Do yourself a favor; get better tools!]

Yeah, I'll just cash in some of those bonds I inherited from... wait... I didn't inherit any bonds.I'm working for less than US minimum wage in a developing country, but thanks for the suggestion. 

[Lunapic is a severely limited online photo editor, not a texture creation tool.]

Except that it does things with sculpts that people can't even seem to figure out how to do with Blender. It's also totally portable; I can work at print shops or cafes if I have to.

[if you don't want to spend the money on Photoshop, that's understandable.]

Someone actually gave me PS. I expected to be a lot more impressed with it than I actually was. It seems like a great tool for duplicating the same work I see everyone else doing in Second Life, if I'm willing to dedicate twice as much time to learing how to use it as I currently dedicate to actually making stuff that sells. The liquefy tool, especially, is nothing like the smart-stretch function that is currently being perfected for my use over at Lunapic. When I explain the difference, you'll probably soil yourself.

[but GIMP and Paint.Net are both free, and Paintshop Pro is only $99.]

I installed GIMP on 3 computers in the US, but it seemed not to do anything. Is it an actual program, or is it just a thing that takes up more space on hard drives, like Blender?  I won't have $99 for anything until I have some kind of account that will allow me to cash out my $L. I'll try Paint.Net, thanks, but I don't actually texture sculpts; I sculpt textures, and the results pay for themselves. If something is not free and fully portable, I'm not convinced that it's going to be any improvement in my case. My competitors are still charging many times what I do for a photoshopped rock that has only an arbitrary relation between the shape and the surface image, and while being seamless, have a pucker at each end, making them difficult to hide simultaneously. My rocks look like a rock except for one pucker, and my only clear production cost other than my time is the 20L needed to load the data, assuming I don't use a pre-existing surface image. Paying more for a better result would make more sense to me if I could see people doing that, but I mostly don't, and that's a gross understatement.

[All three of those options are full-featured image creation programs, more than suitable for high quality texturing at great speed.  Lunapic simply isn't.  The amount of time you've been costing yourself, and the limitations you've imposed on yourself, by using such an under-capable tool as Lunapic are staggaring to think about.  It almost makes me weep for you.]

There's no lost time; I have multiple tabs open in 2 browsers, so the edits go through on Lunapic while I'm doing about 5 other things, including sculpting and manually correcting SLM delievery errors. Also, my time isn't worht sh## in the first place, since my M.A. is in Music Theory. The admin at Lunapic also responds to any problem I report within a matter of hours, and he actually puts in new functions at my request. Does Adobe do that for you?

[if the likes of Lunapic are what your texturng experience has been limited to thus far, then I have to say it makes sense why you'd be looking to options like resizing to solve your aliasing problems.  99.99% of the tools we'd normally talk about for preventing such problems (as well as for preventing tons upon tons of other problems) simply don't exist in applications like Lunapic.]

True. And I can always buy a Sherman tank if I need to kill a cockroach. Or not.

[Just so you know, I've been posting on this forum almost daily for the past seven years, and you're the first person I've ever seen say they've been using Lunapic for texturing.  I'm not even sure I've ever even seen it mentioned at all, come to think about it.  Unless you're a very dedicated massochist, get yourself something proper.]

I'm dedicated to cost-efficiency, and I'm winning. I'll go out on a limb and suggest that a lot of people buying things they will never learn how to use well enough to justify the cost are not winning, especially if they're sinking even more money into SLM listing enhancements and in-world shops. If you've seen any of my stuff, I think you'll be shocked at how much of it truly does not actually suck, especially considering that I'm using 128's and whatnot in a few cases. I don't size up everything, true. But only because sizing up is no greater a guarantee of a better result than is buying a more expensive or learning-intensive tool that all cool kids seem to have to have this year.

 [irfanview is also free, and it does batch conversions.  If a stand-alone converter is what you want, Irfanview is considered by many to be the best one out there.]

This is an excellent suggestion.  Will try it.

[Just about any full-featured image editor will, of course, also do batch conversion, including the aforementoioned GIMP and Paint.Net, which are both free.]

GIMP has failed to do anything on any computer where I have either installed it or had it installed. I have begun to think it's pure mythology, like Blender and Wings. Tech people never cease to amaze me with how much they take for granted in terms of the million hoops that need to be jumped through in order to get anything to actually work. They also never cease to amaze me by making unavailable things that actually work pretty well. To his credit, Cel Edman at least did not actually eradicate Sculptypaint 092, but simply hid the link. Most impressive among techies is the small amount of average creative output per level of acquired technical skill. I'd rather size something up before I cut a circle into it than take out a loan and go to a technical institution where they spend months getting around to the question of how to avoid size-up, and then not really answer the question, thanks.

 [i've never seen that happen before.  It's more than likely a symptom of the tremendous limitations of the specific software you've been using. ]

I've seen it with multiple editors on multiple browsers on both Mac and PC. Possibly it's a symptom of not having equipment I can't afford or afford to learn to use. I don't mean to disparage you personally, but it's mantra I hear repeatedly from people who seem to be able to buy and learn to use basically anything. If you are productive, I can tell you that many others are not. If someone is driving half as many straight nails as I am, it makes me not want to listen when they tell me to stop using a brick and buy a shiny new hammer.

[it sounds like the transparency in your PNG's isn't being generated in a way that the SL uploader fully understands.  The PNG format supports multiple forms of transparency.]

In other words, SL doesn't really understand PNG anyway, because SL only uses TGA, and doesn't correctly convert some types of transparency data when PNG is converted to TGA.  I'm not going to stop using a free portable editor with a responsive, interactive admin just because SL can't get it's sh## together to save me one step when I load transparencies. The conversions have so far been instantaneous with Converthub; the new desktop icon appears faster than I can get my finger up from the button.

Since we're already disagreeing about things, though... are you at all interested in trying to defend Google Chrome on the pixel number issue?

[Shelby Silverspar]

The wonderful thing and at the same time terrible thing about SL is that anyone who wants to be a graphic artist can be. Pretty much the same as in RL where anyone with a computer and a copy of Photoshop or Publisher thinks they too are a graphic artist.

Good habits are easier to learn than bad habits are to break. I would much rather teach a newbie PS user than someone who has years of bad habits built up. Those are the hardest to break.

Josh, you sell rocks. I can see why Photoshop would underwhelm you.

As an aside, the word "upsize" when refering to digital graphics should be replaced by "ruin your graphic" for all intents and purposes. Start with the canvas size you need, never upsize UNLESS you are using an application that is designed especially to do that. There are a few. A better idea is to just start with a decent sized canvas.

There is no point in discussing anything graphic with people who already have their minds made up and have nothing but distain for the tried and true, industry standard and used by the masses applications like Photoshop, let alone Gimp, PSP or Blender. Just because you don't know how to use those applications does not make them useless or bad or not worthy of your attention.

Photoshop is an amazing program that teaches me something new every day, even after using it since version 2.5 or so. Granted it does have a steep learning curve but the freedom and control it gives you are worth every second. Given the quantity of teaching tools and tutorials available for Photoshop these days, self-starters can become pretty proficient mighty quickly.

Chosen, my hat is off to you again and again. You take on both the beginners and the jaded oldbies with grace and determination and give out some of the best advice, tips and tricks I've ever been privy to - both at university and in my professional career (which spans over twenty years). Bravo! You are quite an asset to this forum. I wonder how many people truly appreciate the wisdom you spread here. Thank you.

 

 

[Josh Susanto]

[The wonderful thing and at the same time terrible thing about SL is that anyone who wants to be a graphic artist can be. Pretty much the same as in RL where anyone with a computer and a copy of Photoshop or Publisher thinks they too are a graphic artist.]

I don't claim to be a graphic artist. I claim to be  guy who knows how to get a smoothly curved line the first time. A graphic artist is apparently someone whose job it is to keep telling me and the other person that it can't possibly work, even when we have seen it work, and seen it work repeatedly. 

[Good habits are easier to learn than bad habits are to break. I would much rather teach a newbie PS user than someone who has years of bad habits built up. Those are the hardest to break.]

As an English teacher, I agree with that. I just don't start students by making them buy Webster's Unabridged and start memorizing it.

[Josh, you sell rocks. I can see why Photoshop would underwhelm you.]

Thanks for appreciating that at all. So far, I think I probably sell the BEST rocks. Of course, if I decide that I want SL to look less like the original print of The Trouble With Harry and more like Toy Story, I'll certainly change my ways.

[As an aside, the word "upsize" when refering to digital graphics should be replaced by "ruin your graphic" for all intents and purposes. Start with the canvas size you need, never upsize UNLESS you are using an application that is designed especially to do that. There are a few. A better idea is to just start with a decent sized canvas.]

I don't think we're disagreeing on that, actually. I think it's a question of with what we are starting. I start with photographs. Other people start with templates, other people start with a blank canvas. As I understood the original question on this thread, it was just a matter of how to make a curved line not so jagged on the edge of an article of clothing. Sizing up CAN and DID solve that specific problem, and would have done so, regardless of what software had been applied. 

I will consider using an amp that goes to eleven when Second Life's speaker system at least goes to 10.

When is that, Nigel?

[There is no point in discussing anything graphic with people who already have their minds made up and have nothing but distain for the tried and true, industry standard and used by the masses applications like Photoshop, let alone Gimp, PSP or Blender.]

My disdain is not for applications that work hard for users. My disdain is for people who expect me to work for applications when I don't have to. Maybe you can sell some violins to the Beijing Opera on the premise that they have twice as many strings as an erhu.

[Just because you don't know how to use those applications does not make them useless or bad or not worthy of your attention.]

How much attention should I continue to give an application that doesn't even open, or an application which, once open, seems to have every single button disabled? If I have to attend graphic design school AND systems admin school to make exactly the same smoothly curved line, but in the "correct" way, you can count me out, thanks. 

[Photoshop is an amazing program that teaches me something new every day, even after using it since version 2.5 or so. Granted it does have a steep learning curve but the freedom and control it gives you are worth every second. Given the quantity of teaching tools and tutorials available for Photoshop these days, self-starters can become pretty proficient mighty quickly.]

Photoshop IS amazing, and I have asked friends who are good with it to help solve specific graphic problems. They have also asked me how I get certain effects they can't seem to get with Photoshop. I believe I am ethically entitled to both use and speak well of any graphics tool that easily and transparently does something useful that Photoshop either cannot do, or does not do in any obvious way.

"You can't do that without Photoshop" is just one of the phrases I have had to add to my list of similar things I have been hearing my whole life:

"You can't grow enough food to feed yourself on a piece of land this small."

"You can't successfully sue your employer without hiring an attorney."

"You can't remove that kidney stone with an instrument of your own construction."

etc.

My standard response has become "watch me".

 

 

 

 

[Eloise Baily]

I do this differently.

I use the pen tool to draw round the edge, then right click on the path and use stroke path. I have my brush set at around 4 pixels and choose brush. Leave the lil selection box in the popup that comes up unticked and click ok. It gives me smooth edges every time, and I can use the path again with different brush sizes to create things like stitching.

It may not be the correct way to do it but it works a treat.

BTW, if you hold alt and click on your last pen point, you dont end up dragging the whole thing around, just the curve you're creating in that movement.