Topic awaiting preservation: Power Distort (Page 1 of 1)

I was heading this way, so I thought I would start a new thread to get my rough draft started. Going to take it slow and in small chunks.

Welcome to Power Distort Part I.

This one starts with Broken China. Yes, one of the very first tutorials I wrote. My first or my second - I can't remember.

Going to be using the Zoom Cube as grabitude reference, but in reverse. So, make your own and invert it. Or just build it inverted.

Pick a colour from it and build a new D-Map using that colour. For mine, just a square on a new layer with Preserve Transparency turned on. Going this route so I can easily pick a new colour from the inverted D-Map and quickly fill that square. Save it and give it a whirl.



In my quick test, Biker Chic is actually 256x256. Because of those dimensions, I used 100%/100%. There are times when you will want to use 200%/200%, but it gets really tricky. For now, use a 256x256 target and use 100%/100%.

Hopefully you can see the correlation.

Pick a different colour from the inverted Zoom Cube, update and save the simple D-Map, and Displace again using 100%/100%. You should be able to see the same correlation. Do it again and again. Keep doing it until it clicks. Then do it a few more times.

Please, do this until it makes sense. I can't stress this enough. Even if you understand the graphic, do it anyways.

That's it for now.
Gotta go.

[This message has been edited by warjournal (edited 10-26-2003).]

[This message has been edited by warjournal (edited 11-06-2003).]

[Frying post to clear topic.]

[This message has been edited by Ruku (edited 10-14-2003).]

You are looking at it. This is the very beginning.

[Frying another post]

[This message has been edited by Ruku (edited 10-14-2003).]

Power Distort I.I

Next little exercise is almost exactly the same thing. Almost exactly.

Just to keep things a little straight, be using the same colour as above. However, the second D-Map is a little different. Instead of the solid colour square being in the middle, it is a little bit up and to the left.



Notice the two red lines. The first red line goes from the center, or 50% grey, to the sampled colour. The second red line goes from the solid colour in the second D-Map to where it grabs from. Now, both red lines are the exact same length and direction. However, their placement is a little different. In the resulting target, the red line starts from the solid colour and not the center. The second red line is transposed based on where the solid colour is in the second D-Map.

I have no idea if what I just said makes sense. This is one of the reasons why I've held on to this technique for such a long time. I'm finally getting comfortable with my writing skills to attempt, but I'm not sure if it's enough.

Anyways, play with it. Get to know it. This is very important concept to understand.

[This message has been edited by warjournal (edited 10-26-2003).]

[This message has been edited by warjournal (edited 11-06-2003).]

(Pardon the numbering in the title. Be fixing this in the final.)

Power Distort I.III

Got the hang of the above? I hope so, because it's time to reverse engineer. But first, let's talk about some numbers.

In the above, I have three different documents:
1. Reference Cube. This is the Zoom Cube, but inverted.
2. Working D-Map. So far, nothing terribly complicated. Just a bunch of 50% grey and a solid colour.
3. Biker Chic, my favorite target. She's a thing with me. Some day I'll give her own page in homage.

All three of those documents share the same working dimensions: 256x256.
Also, only been using 100%/100% when it comes time to Displace.

Earlier, I mentioned that sometimes you will want to use 200%/200%. Well, that time has come. Why higher percentage? That's a bit trickier to explain. There are two reasons, but for now only be concerned with one: the reference cube is now has the dimensions of 512x512.

Did you catch that? We'll be using a reference cube with dimensions 512x512. Since the reference cube is doubled, sampled colours will be "halved"; and, to compensate for that, the percentages are doubled.

Let's run through a simple example of this exercise.

Upper-Left
Our target with some stuff tossed in. Instead of a silly square, be using a silly circle. The green circle is where we want to pull to, and the red circle is where we want to pull from. If all goes well, Biker Chic's face will be "pulled" into the green circle.

Upper-Right
This is the reference cubed. Remember: 512x512. The center is marked, and the line in Biker Chic as been dragged in and aligned to the center. The colour at the end of the red line is sampled. I got R=179 and G=173. That is the colour that will be used in the simple D-Map.



Lower-Left
This is the final D-Map using the sampled colour. Pretty amazing, isn't it?

Lower-Right
This is what that simple D-Map does with 200%/200%. Oh, man - that is too cool! That is, believe it or not, pixel perfect. Now, there are some funky jaggies around the circle. This is due to anti-aliasing in the D-Map. Those little shades in-between will do that to ya. Sometimes it's good, as we'll see later, but sometimes it's bad.

Practice doing this.
Pick two spots in your target, and get one to "move" into the other.
Once again, I can't stress this enough.
Again, I can't stress this enough.

At the very bottom of all of this, we have my favorite equalities: 128px = 100% = brightnes 255. Once you understand that and how to manipulate them, it's all good. Believe me, once I understood this explicitly, I was very happy - drunk with power, even.

play.fiddle.learn

[This message has been edited by warjournal (edited 10-26-2003).]

[This message has been edited by warjournal (edited 11-06-2003).]

Could you please make a example where this would be useful.

I don't understand your request, Arthurio.

Right now, it's all about explicit understanding, and that is always useful. The above is a large part of my process when I'm answering somebody's question on how to do something with Displace. With out the above, I wouldn't be able to do the things that I do with intent. Maybe someday, people won't need me to explain how to do specific things. Rather, they will be able to come up with the techniques on their own as they need them.

You know, teach a man to fish...

When we get farther down the line, there will be readily apparent uses. For me, it's mostly 3d related type things. But there are other uses. You'll just have to wait. Hopefully next week this whole thing will unfold into something beautiful - and useful.

Jumping the gun a little bit on this one. I'm naughty like that.

Power Distort II.I

Are you getting the hang of moving things around any which way you like? I hope so. While moving things around as such is cool, let's add a little extra something.

Across the top:
A really simple D-Map with two solid colours. That ugly shade of green grabs from blow and to the right. That ugly purplish whatever colour grabs from above and to the left. See? Not too bad if you've been playing with previous exercises.



Across the bottom:
For me, the next logical question was: what if you bridge the gab between two colours? So I tried it and got some major coolness. Look at that! The distortion is thusly "bridged" as well. Very major coolness.

Quick note:
In this example, I used 100%/100%. I took it easy on the percentages to help keep the distortion to a visually manageable level. If I had gone much higher for the colours I picked, the distortion would have "over-lapped", and it can be confusing to the untrained eye.

I can't believe it. I'm so excited that I'm at a loss for words. Is anybody out there starting to see the potential? Starting to see the level of control?

Oh, man!

You've got to try it for yourself. I'm not saying that in a teacher-to-student kind of way, but in a see-it-for-yourself-to-believe-it kind of way.

I'm figuring on two, maybe three, more little lessons to be done with this tutorial. Until I get to the next part, play with it until your brain oozes out of your ears. It builds character. Seriously.

play.fiddle.learn

[This message has been edited by warjournal (edited 10-26-2003).]

[This message has been edited by warjournal (edited 11-06-2003).]

I'm dying over here. I'm just so excited over this. All those years ago, when I first got into Displace, I came up with this stuff. You folks aren't just seeing a technique, but the actual controlled experiments that I did to reach my level of understanding. Not just a technique, but a chunk of my history at it's very beginnings.

This is my exact journey - literally step-by-step.

I feel so... I don't know, but I'm sure the French have a word for it.
My heart and soul is in Tech-Slop, but I feel like more than that is in what's going on here.

quote:

---

I feel so... I don't know, but I'm sure the French have a word for it.

---

What do you mean Well, let me suggest the word "febrile", or the phrase, in french this time:

'je me sens tout chose'

It sounds like children's words ( and thus can sound humoristic ) and means that you feel some sensations ( generally it's pleasure or pain ... or both ) in your whole body a bit like when you've got the goose flesh but stronger and with a naughty touch.

Does it fit your feeling ?
In fact, your feeling may simply be nostalgia. But the end of your post suggested me my previous idea.

Mathieu "POÏ" HENRI

[This message has been edited by poi (edited 10-18-2003).]

We are not worthy! We are not worthy!
Great lookin tut. WJ. Keep it up! This is gonna be sooo cool!

---
"The Mystery of Life"
Vol. 841, Ch. 26
"All life begins with Nu and ends with Nu. This is the truth! This is my beleif! ...at least for now." - Chrono Trigger
><>

Beautiful stuff here warjournal =)

The potential is quite awesome.

Poi, it like the first thing you said, with a bit of nostalgia. I used to have a math teacher that used to get the feeling when talking about certain proofs. Whenever he talked about natural log of e^x, he would get overly enthusiastic. (Maybe it was the derivative of the natural log of e^x?)

I'm not going to lie about it - I do get very excited about this stuff. Man, too much stuff to say.

(Although, I must admit that I'm also feeling a bit of remorse as well. Technical Slop has almost come full circle. The time is drawing near, folks.)

I might jump the gun again with the next part. Here is a big hint from an older post from other forums:

quote:

---

Imagine you have two quadralaterals of any shape. The corners of both can be anywhere you want. Now imagine being able to transform a texture on one into the shape of the other. With several sets of such quadralaterals, you can seamlessly wrap a texture to corner on a cube.

---

With what I've provided so far, figuring it out shouldn't be too hard. If someone does figure it out before I get to it, I'll be *very* happy. Why, I'll be moved to tears.



[This message has been edited by warjournal (edited 10-18-2003).]

Power Distort II.II

Alrighty, then. We've got moving solid chunks around. We've even got briding between two solid chunks. What's next?

I know! Let's do a cube! It's not that hard. Really.

I picked four colours to be in the corners. Both of the upper colours being the same is just a bizarre coincidence. No particular reason other than clicking twice on the exact same spot. From those four colours, we get two gradients: one gradient going top-to-bottom, and the other gradient going left-to-right. Then slap one more gradient in there in a Layer Mask.

In each corner is a set of numbers for those colours. The first number being Red and the second being Green (Blue = 128 all over the place). In the lower-right, the hidden colour is R=158 G=100.



See? Not too difficult. Let's take it for a spin.



And there she is with 75%/75%. Again, I took it easy with the percentages to keep it visually manageable. Had I gone much higher, I would have been "out of bounds", as it were.

And, again, the red shows "grab from" and the green shows "pull to". The white lines are the lines grabbed from the reference inverted Zoom Cube.

Now that right there is major cool. It means you can take any quadralateral of a picture and transform it into a square! Even though I was a bit sloppy with this example, it is possible to be as pixel perfect as you want it to be.

Man, this is some exciting stuff! I scoff at people that doubt the power or control of Displace! I scoff! Scoff, I say! Ha!

Just a little bit more to go, and I'll be done with this monster technique. In the meantime, play with it. Take it out for dinner and get to know it.

play.fiddle.learn


[This message has been edited by warjournal (edited 10-26-2003).]

[This message has been edited by warjournal (edited 11-06-2003).]

WJ - This all sounds really interesting, but I can't see any of your sample pics... Given some time and my own image, I'm sure I could get the effect, but it'd sure be nice to see what you're doing here...

warjournal: One thing that misguided me was the 128 blue. From the few tests I've done, the blue component is useless and not taken into account. Another thing that may help the 3D coders and artists out there is to really consider a Dmap as a normal map, cause that's exactly what it is. Well, it helped me at least. The red component stands for the X axis, while the green component stands the Y axis.

Now that I have the mental image of a normal map, I really understand how to control the effect of a Dmap. The problem is to figure how to create the correct Dmaps/N(ormal)maps

Mathieu "POÏ" HENRI

Bodhi, I was borrowing space and bandwidth from DB. Well, DB and his server went away for the time being. I'm working on getting images to another server.

Poi, I never really thought about D-Map and normal maps. Pretty good analogy if you are familiar with normal maps. And I almost always fill Blue with 128. It helps me see the R and G channels better when looking at the composite.

I got the images back on my other server. You should be able to see them, but this server is the tempermental one. My fingers are crossed.

Power Distort II.III

Have you been playing with it? I hope so. I have a tendency rely on the foundation so I can move things along a little faster.

So we took a funky quad and turned it into a square. Pretty much just like Crop Perspective. But we can take this a step further with Transform Distort.

In the previous D-Map, all of my gradients were on their own layers. So, selected them and hit Copy Merged (ctrl + shift + c). Pasted that into a new layer and hid the other gradients. On the merged that just got pasted, hit Edit > Transform > Distort, and give it some wiggle. Ended up with something like this:



I took it easy on the left corners - not much difference there. The corners on the right, I moved a bit more.

Save it and play. What you should get is a funky quad turned into another funky quad, but with the white lines the same length and direction as the square example. Something like this:



If you compare the white lines in this example and the white lines in the previous, notice that they are indeed the same direction and length. However, they start at the corners of the D-Map (which got distorted). Does this idea sound familiar? I hope so. That's what building a foundation is about.

Again, I was a bit sloppy with the lines. And, again, Red is "grab from" and Green is "pull to".

Like I said earlier, you can take one quad and distort it to another quad. The corners can be anywhere you like, and you can be as pixel perfect as you care to be. Although, you might not want to get too crazy with the corners. If you get too crazy with them, chances are you will get jaggies that are hard to deal with. I still haven't figured this out to my satisfaction.

Play with it!

Only one more to go.



[This message has been edited by warjournal (edited 11-06-2003).]

Hey WJ, that's excellent stuff man. "Distort" really does rock!

WJ... go work for Adobe.

Yes, I have been slacking a bit. I'm sure some of you folks know how family life can be. But the extra time I've been taking on this is a good thing. That is, it's more time for you folks to play around with and understand the ideas presented so far.

Again, I'm relying on the foundation to kick things up a notch. If you are not familiar with previous ideas, chances are you will get lost.

Power Distort III.I

I've shown that a quad can be made into a cube, much like Perspective Crop. Also shown that a cube can be turned into a quad, much like Transform Distort. With that combo, you should be able to take a bunch of quads and stitch them back together in new ways. Say, like the on three visible sides of a cube.

But there are problems with that route. One of the nastiest problems is hard edges. That is, no real blending between the chunks. However, I've shown that the distortion in Displace can be blended. Remember the first gradient example? Well, that idea can be expounded to include quad chunks that you stitch together in a D-Map. Then, when you use Displace, the distortion between the chunks is blended as well. All it takes is a little bit of Gaussian blur.

A cube for an example? Sounds good to me.

Upper-Left
This is a cube I made with Filter > Render > 3D Transform. I turned the background off so that the cube is all by itself with transparency around it. I also used the Magic Wand to turn the various sides into it's own alpha channel. So, just from that one cube, I have 4 different alpha channels: top, left, right, and composite.

Upper-Right
Here are my sets of quads with the lines drawn in. Red is "grab from" and green is "pull to". I was going for three diamond shapes that are the same size, but I was a little sloppy. The lines are on a seperate layer, of course, which I drag-n-dropped into my reference cube. Then I got busy collecting R and G colour values for the various corners.



Lower-Left
With my values collected and written down, I made the gradient cubes and used Distort to move them into place. Each one of these chunks, again, on it's own layer. I also darkened the background a tad for illustration purposes, so don't mind that.

Lower-Right
This is the finished D-Map. I copied each individual chunk and merged them together in one layer. Then I gave it a bit of Gaussian Blur to make the edges smooth and sexy. Remember I mentioned alpha channels? Well, turns out that I didn't need the sides, but I did use the composite alpha channel. I copied it and gave some Gauss and Levels to smooth out the corners and tighten it a bit. Used that for a Layer Mask and tossed in a 50% gray background underneath.

Saved it out and gave it a whirl.



Tada. Pretty snazzy, huh? I even went so far as to paint in some shadows and a bit of specular. Remember those alpha channels that I didn't use? Well, I could have used them to do some more chopping to add a reflection on the surface that the cube is resting on. But I'm perfectly happy with my quick paint job. Although, she does have a bit of an Elvis Presley snear going on.

And there you have a simple example of Power Distort. Other uses? Well, I've used this technique to give cars and other vehicles a better custom paint job. I've even used it to drape patterns on furniture. A few other things that I don't want to get into just yet.

There is going to be a Power Distort III.II, but you are going to have to wait for the "official" tutorial. This gets into even more heady stuff and is a seque for yet another series of tutorials that I've been planning on writing.

Until then, you know what to do.

********

Ruku, that is the basis for that one D-Map that I made. Once I had the above cube, I just added some random Ajustment Layer to mix it up for the extra chunks. Makes sense?



[This message has been edited by warjournal (edited 11-06-2003).]

quote:

---

Tada. Pretty snazzy, huh?

---

So just what are you trying to say here??? <lol>

Fucking amazing is what it is. and you can quote me on that =)

Ayup! I concure with NoJive.

It's a work of art WJ, a work of art. Most excellent.

warjournal: I don't remember to have already seen the result of your 3D cube Dmap. But I have no alternative but to second NoJive's "Fucking amazing" exclamation.

I'll be back with my own cube(s) in few hours

Mathieu "POÏ" HENRI

:sigh: It warms my heart and cockles. And to think that I've been hanging on to this since Tech-Slop was a gleam in her daddy's eye. Can you see why I waited for my writing skills to improve before unleashing this monster? I just might have to celebrate with a new sig.

I've given it some thought, and I will probably give a sneak peak into the next part(s). Not much, but enough to get you folks thinking about the next level (which I already hinted at in my last reply to Ruku). Then top it off with an observation similiar to the final observation in Reflection Maps.

Poi, I can't wait to see what you come up with.

Here comes my first dMap cube.



I've spent more time to find a picture to put on the cube than doing the dMap itself Finally, the 'texture' is a picture of me, doing my 'terrorist' face with my hairs unleashed and a ~1 week beard.

I'm thinking to a way to convert a b&w elevation map to a dMap. Using Filter > Other > Custom with an offset of 128 plus a simple kernel [ -1, 1 ] ( horizontaly for the red component of the dMap, or vertically for the green component ). Then, some Levels to colorize the resulting forward difference map in red or green.

Mathieu "POÏ" HENRI

[This message has been edited by poi (edited 11-02-2003).]

MMM is see a whole night off experimenting with this somewhere the next in front of me.... woohoo.

__________________________________________
"Art has to be forgotten. Beauty must be realized."
Piet Mondriaan

I can't see your image, Poi. Suck!
With a b&w height map, I use either Emboss or Lighting Effects. If you come up with another way, I would definitely like to hear it.

Sneak Peak 1 (hopefully you can see this)

Once you get the hang of stitching D-Maps together, it's not that hard to add a little extra something. I think of the extras as secondary motion. I've mentioned this before.

Crappy example across the top just for Ruku. I redid it because I wasn't happy with the lines in the old version. It's just a little extra chopping for the backfacing and an Invert Ad-Layer. No big deal. Of course, you could sit down with pencil, paper, and a calculator to really make is fancy. For me, something along these lines is enough.



Across the bottom is a better example of secondary motion. I made a little divot with some fancy-smancy Gauss, and used the result to mask a Curves Ad-Layer. Like an idiot, I left the background way too dark. But I did bother with some decent finishing touches.

Gotta go.



[This message has been edited by warjournal (edited 11-06-2003).]

warjournal: my host have no direct link blockers and I can see the image in all the browsers I have. Whatever, you only miss a dull cube.

The round punch of your last cube is interresting, not to mention the lighting effect. I guess using a sort of environement map which is Displaced and set in HardLight mode gives a great effet.

The idea of my technique is to use a [ -1, 1 ] kernel with an offset at 128 to get the derivative on the X or Y axis of an elevation map, color the derivative maps in red or green and mix them to get the final dMap. Is that a little more clear ? I got some interresting results, but not as good as I expect at the moment. At least, I think the technique is mathematically correct, and I could make an Action to automatize the creation of a dMap from an elevation map.

Mathieu "POÏ" HENRI

:: breaks down crying:: *sniff* Its beautiful...

[And wow I haven't been here in a long time.]

---
"The Mystery of Life"
Vol. 841, Ch. 26
"All life begins with Nu and ends with Nu. This is the truth! This is my beleif! ...at least for now." - Chrono Trigger
><>

Many thanks to Mr. You-Know-Who.
Images uploaded to new server and posts edited (again).
To celebrate, I'll have a little extra something in the coming days after I post Sneak Peak II.

Poi, I can see your image now. You didn't take a route that I was expecting, but it looks like you nailed it with absolute pixel perfection. Kudos.

warjournal: To get pixel precision, I cheated a little. I duplicated the dMap layer, blurred the underlying dMap several times ( to cancel the interpolation at the edges due to the Select > Transform Selection > Distort ), merged all dMap layers and finally applied the mask of the original cube to it to remove the blurred edges. And voilà.

BTW, with PS CS better 16bits support, I support it'll be easier to get pixel precise dMaps. In my test, 2 faces of the cube had to pick some pixel further than 128 pixels so I've had to divide the opacity of the 3rd face by 2 with brought a little loss of precision.

Mathieu "POÏ" HENRI

This is something that I avoid talking about. As you read through this, don't be surprised if I sound a bit angry.

Sneak Peak 2

Even for a sneak peak, I'm not showing quite as much as I want to.

As we saw way back in the beginning, you can pick a colour from a reference cube and know it's direction when you hit Displace. The opposite is also true. That is, you can start with a colour, then Select > Colour Range in the reference cube to see where that colour with grab from. This is good.

Now, there are a lot of tutorials out there about wrapping a texture to a face or something. The majority of those tutorials are just flat out wrong. Yes, they are wrong.

One method to show that they are wrong involves the above Colour Range trick. Unfortunately, I don't have the strength right now to run you folks through it. If you are smart enough, and I'm sure you are, you'll figure it out.

Another method to show that they are wrong is Power Distort. Luckily, I just happen to have a face to illustrate this.

Upper-Left
This is the face that I chose. If you are using one of those other tutorials, chances are you would be recycling the colour information in the photo. Right? Thought so.

Upper-Right
Oh, man. I stitched together eleven funky quads to get a partial D-Map. It's not the whole face, but it's enough to get my point across. It is a lot of work, but the end result is far superior to recycling colour values from the photo.



Lower-Left
This is the Red channel of the partial D-Map.

Lower-Right
This is the Green channel of the partial D-Map.

I'm sure you can see how both of those channels are a far cry from the values in the photo.

I have tried this D-Map and it works very well. Although, I haven't gotten an example that I'm entirely happy with. That is, I have yet to find a texture or something that looks decent when wrapped. I tried grids and a few other things, and they do wrap well, but they don't look that great for one reason or another.

While most of the time recycling colour information is bad, there are times when it works. A certain someone did a fine job of this. When I get to the official tutorial, I'll get into this.

That's about it from me.

play.fiddle.learn

I just played with the face provided by warjournal and tested my method to convert an elevation to a displaceMap. I made the elevation map by hand by drawing some white shapes and slightly erasing the areas supposed to be further. The whole thing is not awesome, but I did it pretty fast. And finally it proves that the convertion from elevation to dMap works well.



Mathieu "POÏ" HENRI

Excellant, Poi. Painting D-Map like you did is something that I'm planning on covering.
Try inverting your final D-Map and trying that. The one that you have is in "contract mode". If you invert it, you'll put it in "expand mode", and I think it will work better.