Change in EXR open from CS2 to CS3 can this be fixed?

Joey - I don't know where to start.

Please start with the first item in the text that you think needs to be addressed and refute the opinion, argument, or example with evidence to the contrary. From there, Wash, Rinse, Repeat. I will openly welcome a point by point discussion on each item.

You still have a lot of history and terminology confused

Respectfully no I don't think I do. I was rendering 3D animations with CubicompPicturemakerin 1988. I was compositing these animations with Cubicomp Sequence using their .a8 channel during that time. 1988 was barely one year or so after Photoshop was even conceived. A couple of years prior to that a professor was teaching me in regards to the use of key channels, or "alpha channels" as he referred to them as they related to the functional operation and communication between a Laird character generator and a Grass Valley switcher and it's downstream key.

you don't seem to see the connections between CGI applications and Photoshop.

Unfortunately there are few reliable connections between modern CGI applications and Photoshop. Thats what I and numerous others during the course of this thread have been trying explain. Photoshop does not behave like or behave well with most modern CGI applications or the products they create. Do they work, in most cases, but not like they should. I provided to you point by point examples how this is the case with files from both Maya and Softimage. Are there connections, sure, can Photoshop be used in CGI, absolutely, if you want to bludgeon yourself with the Photoshop Paradigm, absolutely there are connections, again I provided direct examples of just such cases.

Your rant on "alpha" is particularly telling.

Well...it wasn't a rant, but you are certainly entitled to view it that way if you wish.It wasn't a technical specification driven example of what an alpha is. I didn't intend it to be that. I was attempting to convey to you how and what we, animators and compositors, use, perceive, understand, treat, and expect from, the Alpha. How the Alpha relates to our world. I don't think you will find any digital compositor that disagrees with my...uh...rant...because these principles have not changed in over 20 years.

You ignored the more common usage of "alpha channels"

Please tell me what the more common usage of "alpha channels" in Maya IFF, SoftimagePIC, SGI, TGA and EXR files are.

and the definition of alpha channels in file formats

Please tell me how many file formats define alpha as transparency, how many define alpha as opacity, and how many define alpha as a mask. In all truth, I don't know the answer to that question.

to rant about how Photoshop's common usage of the term and adherance to file format standards

Please tell me why a Softimage rendered PIC file and/or Maya rendered IFF,  when read into Photoshop and saved out of Photoshop, are larger in byte size when then they are saved out of Photoshop with absolutely no change to the file and saved to the same format they were sourced from?

does not match the imagined usage in your industry niche.

Are you saying that I am imagining how I use and expect to use the images our software creates?  I'm stunned. I've rendered from Cubicomp, Softimage, Mental ray, Maya and Max, and I've composited in Sequence, Softimage Eddie, After Effects, Max Video Post, Wavefront Composer, Discreet Flame, Shake, and SoftimageFX Tree. Did I imagine my use of these apps too?

But then you did say that OUR INDUSTRY niche imagines how we use these things, didn't you? All of us. Thats cool. We all get it wrong I suppose. Not just me. I feel better now. I'm in good company.

You also made a lot of statements about alpha channels which are wrong because you ignored the problems of premultiplied color and straight color.

Premultiplication. Sure, we can discuss this. Premult is pretty easy to understand actually. I don't remember the actual math anymore, and the concept, the whole multiply the color at the antialias edge using alpha as the multiplier thing so that black(the black in the RGB plate under the rendered object) is not added to the comp at the antialias perimeter is well,.... boring.

But suffice it to say you multiply the anti-aliasededge color in order to cancel out the background black rendered into the anti-aliased edge. When using the pre-multiplied image with the standard alpha you don't get a black ring in comp because you canceled out the black in the anti-alias edge as it relates to the alpha's corresponding anti-alias. But you see I don't normally have to worry about those details.

Its like this. Most 3D apps, I'll address Maya and Softimage specifically because they are the apps I use now, render in one of two states. Premult or no premult, some call this straight. Soft and Maya rendering both allow you to change this state and both are set to premult by default and most compositors use this as the default if they are going to comp these images in Shake, Nuke, or whatever. If I bring an IFF file and a PIC file both into Shake, comp with an Over node, I can swap out the IFF and the PIC to the Over node input till the year 3000, they will both work identically with no black alpha ring, because the Shake Over node knows how to handle the rendered premult edge properly. It was the same with Composer, Flame, Eddie, etc. I think in Eddie you had to throw a switch to force premult computation, but it supported it perfectly. Even Cubicomp did 20 years ago.

The point is that premult is either on or off, and all 3D applications support Premult computation fine. So why can't Photoshop support it correctly? Why is it that when I render an object in Maya with premult on and take the rendered IFF file and Place it in Photoshop it has a black ring around it when Shake handles the premult image just fine? Why do I have to go back and rerender in Maya with premult off in order to get a straight image to use with Place? The assumption is that Place is designed to work with straight by default, is this true?  I don't have a clue, but I know my Maya image is rendered with premult on. I also know I have to rerender with premult off to make the image composite correctly in Photoshop. Whats up with that?

Is it the IFF reader that messes this up? Is it Place? is it the Layer? Why doesn't Photoshop use premult by default like every other high-end CGI compositing app? Lets examine this in more detail.

OK, So All I have here at the moment is XSI and Photoshop CS.

I render two images from XSI, both are the same camera shot, single frame of a red crcle over black.

One image is with premult, one is straight.

The premult image is 21 KB large.

The straight version is 15 KB large.

So lets benchmark this, we'll open SoftimageFXTree and create a comp.

Load both images as file inputs

Create  an over node.

Create a mask shape and make it full field white

Connect the white field to the background input of the Over node.

Connect the premult image to the foreground input of the Over node.

Open the viewer and look at the result.

By default Over is created with Background type set to Pre-multiplied.

Since we have the premult image connected to fg and premult as the process method, that is the method and image type match, there is no black ring.

Now if I change the background type in Over to Not Premultiplied,,,,oh my

There is a black ring now. Look familiar...youbetcha...but we'll touch on that in a bit

Now lets connect the straight image into Over fg disconnecting the premult image from the tree

Odd. It renders just fine, no black ring. Hmm...no whats up with that?

Change it back to premultiplied, again no change, no black ring... curious?

Well, whats going on here is that the image, when rendered from Softimage with premultoff, renders the object edge so that there is 100% color not mixed with background black. In essence the image RGB looks like it is aliased as opposed to anti aliased, but closer inspection reveals that what has really happened here is that anywhere the alpha's antialias is rendered at non-zero the RGB anti alias is replaced with full color, giving the comp solid color to blend in the comp and the appearance of an aliased edge. It doesn't matter if we switch Over's background type setting back and forth, between premult or not premult types, the result is visually the same.

So on to Photoshop.

We bring both images, the straight image and the premultXSI images into Photoshop CS.

Place white layers below each

The premult image has a black ring around it

The straight image doesnt.

So why is that? What this presumes to tell us, at least as it relates to XSIFXTree, is that the premult image in Photoshop is being computed as not premultiplied because it is the same result we get when we set the Premult image in FXTree to be computed using Not Premultiplied in the Over node.

So we attempt to change the layer blend mode. Normal, Screen, Overlay etc.

Hard light, Vivid Light, Pin Light and Linear light seem to help, but are these safe for the color in the rest of the image. We want a direct over result, we don't want the rest of the image to be affecting incorrectly.

So blending doesn't really help or doesn't appear to be safe for what we want to do. Normal doesn't cut it.

Is there a true premultiply processing feature in Photoshop? I can't find it.

So after I render 3000-4000 frames of a video animation, comp it and send it to video the client comes in and says "i'd really like to make a montage now of alot of the elements in the animation for some print stuff, can we do that?"

Sure I say, so i load  Photoshop and start bringing in premultiplied rendered foreground elements that I just comped in Shake and they all have black rings around them. So I have to go back to maya or XSI and rerender them all, just for Photoshop. Not for FXTree, not for Shake, not for even After Effects, but I have to do that for .....yep...Photoshop.

So please, tell me again exactly what I ignored regarding the problems with premultiplication?

The fact that you ignored all the corrections and explanations already made in this topic is also pretty insulting to everyone involved.

I did not ignore any of the corrections and explanations already made. I read them. The whole thread. I don't accept some of them. I disagree with some of the assumptions being made about CGI and the work that we do. I don't accept the excuses being made here regarding what can and can't be done to make Photoshop better. I don't agree with these conclusions, because ...after all....we've been told that the way we use these images is...how did you put...imagined?

Truth is, we are the ones who should be insulted. Personally I'm more amused at this point. I've only got several hours of finished animation in my portfolio from the last 20 years, many animations of which had as many as 20 or 30 comped layers and foregrounds(those broadcast news folks...they love to see hundreds of little floaty shiny things in their graphics).  Lets see, 30 frames x 60 seconds x 60 minutes x an average of 10 layers......if i discount the background layers .......that easily a million images I've composited per every hour's worth of portfolio. I think I know how an alpha channel is supposed to work.

Then you rant about file formats with results that don't match any bug reports,

Consider the examples I provided as a new bug report. Seriously, break out your copy of Maya, fcheck.exe, Softimage, etc, create the test images yourself, follow the instructions I laboriously provided. If the results are different tell me how I screwed up. I can take that. But if my examples show what I described, please log the examples as bugs and lets have a discussion about what I have described to you. You did after all ask for someone to provide a good reason.

Photoshop is a vital part of many CG workflows,

As I pointed out in detail. I've worked in Print, Television, and Animation. At one point in the 90s I worked at a television station where we used Photoshop to generate text elements for our Quantel Express. I think I'm well aware of the different workflows.

and aside from people not understanding file format specs or the word "alpha channel" there are almost no problems reported.

So how do you explain the complaints in this thread? We're trying really hard to share with you the problems we are facing. If there are no other problems than ours being reported regarding Photoshop, then Adobe's plate must be really open to listen and accommodate our needs. At least from your account of it, it sounds like we are the only ones complaining.

Again, I encourage you to sit down and replicate the examples I have provided you. I think you will find these things to be illuminating. And don't discount the references to the way we used to do it in Matador or other CGI applications. They may be prehistoric by todays standards, programmed in Fortran and all, but there is alot to be said about the features they had which excelled in their simplicity. We are seeking the same thing from Photoshop.

The few file format problems reported are due to corrupt files causing crashes, or really obscure variants of file formats not working as expected (ie: CCITT images over 30,000 pixels across, 13 bit signed integer pixels not loading, etc).

Once again you need to review the file issues I've illustrated here. The byte size differences, the difference in display between Photoshop and fcheck etc. Photoshop is creating non-standard versions of some these formats. They appear to work, look like they work most of the time, but there are artifacts being displayed in some of the files being generated from Photoshop, the file sizes are wrong or inflated beyond what they should be. You seem like the sort of person who really wants to adhere to the specifications of these formats so I think this would be a serious concern or you.

Yes, you are much too close to the trees to see the forest.

Yes you are right. I have been in this forest for a very long time. But I'm not lost and I've seen some of these tree grow up from little saplings. Some cut down from shear corporate stupidity and others clear cut in the wake of shifting industry winds.

yet we do take the time to learn about and help each small area.  We may not always do things exactly the way you envison them -- because we're looking at a much bigger picture and trying to help a lot more users than just you.

Funny, I don't feel like Adobe is taking the time learn about my area. If I'm mistaken please correct me, but in this response I've been told that the argument i provided you, with excruciating detail and long comparison of what we expect and how other apps did it, is a rant?  I'm further told in this response that I have a lot history and terminology confused, I dont see the connections, oh yeah I rant some more, I ignore common usage of alphas, I'm wrong, I'm insulting, and oh yeah, I...we in this industry, have an imagined understanding of it. Forgive me, but we're not feeling the genuine concern that Adobe is really interested in our plight. In fact....at least for myself, I feel belittled by this whole process.  Whether its pointed at me or other people on this thread who were trying very hard to convey their problems with the software. I know what they are dealing with, been there done that. They're not making this stuff up. Its not imagined. And when you have to fight with these inconsistencies not just once, or here or there, but for hundreds or thousands of images, you start scratching your head wondering, why is this workflow so poor. After all, making an animation  is not like creating a poster. There could be tens of textures involved and hundreds if not thousands of rendered frames.

I went through this exercise to share with you some things which I felt had not been discussed so far in this thread. The EXR thing is only one problem, but ultimately its not the worst of these problems. I was trying to illustrate how these other problems, while related to EXR, are part of a much bigger workflow problem. I went through the enormous detail and length that I did to provide you a serious and fair criticism of whats wrong with Photoshop for us. You did after all ask for that. Twice. I did NOT want to respond with "photoshop sucks", because it doesn't but its a far cry from a good fit in this industry. I've seen my applications of choice, Matador, Aurora, gFx, etc all fade into history. I use Photoshop because there is little to nothing left that can do what it does and do it across CGI and Print. Just because Photoshop is king of the hill now and no one is trying to make a custom CGI app anymore doesn't mean that Photoshop has it "right". It just means it survived, mainly because of it price point. My first seat of Matador cost us almost $15K back in the mid 90s. There are thousands of people who own licenses of Photoshop. Only a handful push the software to its limits. We are a large part of that handful.

I have had these arguments regarding Photoshop for years. There are some people who work in this industry who think PS walks on water. "PS is the industry standard", "PSD files are the industry standard", and all that.  I don't want to deny them their loyalty, but in many cases they are wrong. It doesn't work like it should, and in many cases they know it but can't explain why. And not being able to explain makes it easier for them to maintain their undivided loyalty. Thats why I wrote this. I wanted to explain why because I know why I have problems with this software. I know what I want it to do that would make it better. Because I have a perspective, it may be a bit prehistoric, but a perspective that most, no 95% of Photoshop users don't have. I've used Matador, I've used Aurora. I've used applications designed specifically for my industry. In a lot of ways they were superior to Photoshop for my needs. For $10K or more they had better be and they were.

BTW - Adobe has a pretty huge presence at SIGGRAPH every year (check the authors for the past few years), and other CG conferences, and image processing conferences, and CG trade shows, and ....

Papers sure, but how many of your developers and engineers show up to talk to Adobe's average user? Its been many years since I've seen Adobe in the Exhibition. 2001 or 2003 maybe. As for other trade shows, I'm not interested in those. I'm interested in Adobe showing up at Siggraph with real people to listen to their users in this industry.

Finally I would ask again, run the examples I provided you for yourself. If you don't have the time find an intern or someone else to do that and get their opinion on whether I provided you a good example. But verify these results against the apps the final images or created for. I'm sure you've got copies of Softimage and Maya, don't you? Shake? No...probably not Shake, ....get a copy of Nuke then. If my examples don't measure up, let me know how. If they do, then lets figure out a way to fix it or at a minimum reduce the number of commands and actions required to do the work. You're a smart developer, I'm not worried about your ability to program the features, but I think you and your team need a different perspective on what people want from this application. I'm more concerned with workflow issues, fewer keystrokes and buttons to push, ease of use. Stronger compatibility with the CGI softwares I use. Preserving my alpha channel from unecessary corruption.

I, and a lot of others in this industry can provide you a lot of insight into what would make Photoshop a real, rooted staple in this industry. We need to do things such as rotoscoping(Matador, that prehistoric beast, did that you know), sub pixel rendering(Matador did that too), along with all the other stuff I already mentioned. We're very eager to share our thoughts. We just need someone to listen.

Joey Ponthieux