Compositing and Post Work - What is it, and why should I care ?

145791014

Comments

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited April 2015

    And to show just a hint of the incredible flexibility you have, below is the same image, but with two 2D colored lights (blue and green) operating off the same normals pass. You can add as many lights as you want, with different colors, etc. It's basically unlimited what you can do.

    And you can even modify the levels, etc., of your normals pass to get different intensities, spreads, etc., of the simulated lights. And you can watch the changes happen in real time.

    Normals2.JPG
    544 x 494 - 42K
    Post edited by JoeMamma2000 on
  • DesertDudeDesertDude Posts: 1,225
    edited December 1969

    Very nice work on the fog effect.

    Thanks evilproducer!

    Thanks also to Joe for explaining how to use a Normal pass.

    +1

    I never knew how that could be used, thank you!

    DesertDude, very nice.

    The key to most of this is just what you're doing...thinking. Trying to figure what you want, and then trying to figure how to separate that into elements and arrange the tools to work on those elements to get what you want...

    Thanks JoeMamma2000 for all your examples. In the past I found it very difficult to find, or at least make sense of information about compositing render passes...aside form rudimentary examples. I thought it was perhaps a "coveted" art(!).

    Looking to push this further, thanks Carrara community!

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited April 2015

    Another Carrara pass that I forgot to mention is the "Position" pass (aka, "World Position" in other apps).

    The position pass is pretty much what it says. The data in the grayscale images in the pass describe the X, Y, and Z coordinate positions of every point in the 3D scene. So when you render your image, you essentially have the position of every point on every object that is visible in the image you rendered.

    Now, that might sound a bit like the Depth (Z Depth) pass, and it is. But while Z depth is only distance from the camera (ie, Z coordinate only relative to the camera), Position is X, Y, and Z position relative to the world of the scene.

    So what can you do with this pass?

    Well, we know that a simple depth pass can be used for fog and blur effects. Effects that are a function of distance from the camera. But the position pass can be used in a volumetric fog effect that takes into account the objects in the scene. So, for example, you can simulate an animation of smoke or fog creeping around the corner of a building. It allows a true volumetric effect, but in a 2D image. Very cool, and extremely useful. While a simple depth pass fog can tend to look fairly weak and lacking depth, using a volumetric fog or smoke can provide a lot more character to the effect. And with most good compositing apps, the flexibility in what you can achieve, even compared to some 3D volumetric fog simulations, are quite impressive.

    Here's the position pass from the simple scene I've been working with.

    Position.jpg
    1200 x 1086 - 130K
    Post edited by JoeMamma2000 on
  • DUDUDUDU Posts: 1,943
    edited December 1969

    Joe, I have a question. I makes since always my renders in targa seq. (and .bmp), I noticed that PS (cs5) didn't want to open the .tga of Carrara whereas on the net one says that PS can open them.
    Is this the same thing for you or, am I the only ?
    I don't really need that because AE and Combustion accept this format, but I find that very strange.

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    One other very powerful and useful aspect of using a volume fog or smoke with position pass data is that some apps allow you to also import 3D light and camera information as inputs to the volume fog calculations. This opens up another world of cool stuff you can do to a 2D image, as if it was a 3D scene.

    But one thing to keep in mind if you intend to use features like this is that the bit depth of your render passes can become an issue. Some of the features don't work very well with only 8 bit/channel data, so your results might be less than desirable. And since Carrara's renderer only generates 8 bit/channel data, you might have some issues.

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    And for just one of many, many, many examples of how all of these and other compositing techniques can come together to produce something that is pretty incredible, here is a link to what is called a "breakapart reel", which breaks apart various visual effects and composites to show you the before and after.

    This one is for a film called "District 9", and the vfx was done for the most part by Image Engine in Canada. The compositing work, along with the rest of the production was really quite impressive. A premise that was total fantasy, but presented in a way that was very believeable.

    I suppose this could also be posted in the other thread about unbiased rendering. :) :) :)

    Image Engine used a biased renderer named 3Delight, and I think it's safe to say that most would consider the results as very believable and realistic. And if you look closely at the breakapart, and the individual steps portrayed the the composites, you can see what it takes to integrate some renders into a real life shot. Color correction, atmospherics, etc.

    And you can also start to see all of the factors that are required for something that is 3D to look "realistic".

    https://vimeo.com/95324453

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Okay, now let's take an image which has everything but the kitchen sink (glow, SSS, GI, post effects, reflection, etc.) and take a critical look at it and see what if anything we want to change about it. And then we can step thru the process to actually make those changes.

    Below is an image that I posted here a while ago. It has reflections (mainly in the floor and in the car body), a Glow post effect on the big LED screen on the wall, gobal illumination (which is fairly minor here since there isn't much diffuse bounce light from the surfaces), and some soft shadows.

    I rendered out passes for just about everything, including object index and distance.

    So let's take a look and see what we can do to improve it.

    Beauty.JPG
    950 x 734 - 81K
  • cdordonicdordoni Posts: 580
    edited December 1969

    Hi Joe,

    I notice a slight checkerboard pattern in this image, which is typical in Photoshop when you have multiple layers with transparency. Flattening the image usually gets rid of this effect ...?

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited April 2015

    The first thing which stands out to me is the girl's dress is very blown out. So typically we'd go in, figure out what's causing it (specular, diffuse, GI, etc.), and modify.

    Unfortunately, I am more and more convinced that there is something amiss in the way Carrara handles its render passes. And I have come to the tentative conclusion that when you take all the component passes and bring them together, the image you get may be slightly different that the "beauty" pass.

    So in this case I took all the passes, combined and blended them as they should be, and came out with an actually nicer looking image, where the dress isn't as blown out.

    Go figure. Sometimes stuff just works and you're not sure why. :) :)

    CarFinalDressBlowout.jpg
    1000 x 773 - 331K
    Post edited by JoeMamma2000 on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Now the next thing I see is a rather annoying big, bright specular highlight on the floor in front of the girl. And if you look at the specular pass below you can see why. There's a big white ball of specular which is taking away the focus on the important subject....the chick in red.. :) :)

    So how do we do something about that? Anyone? (Hint...ISOLATE).

    Okay, we isolate that specular annoyance, while leaving the rest of the specular highlights alone, because they look cool.

    In your compositing program you load the specular pass, and make some sort of "mask" which will either delete that specular completely, or tone it down a bit.

    If you're working in PS, just add a white layer mask to the specular layer, and using a soft, black brush just brush out the area around the big white specular highlight. You can vary the softness and opacity of the brush depending on how much of the highlight you want to remove.

    Attached is the resulting specular pass after brushing out most of the highlight, and the resulting image without that highlight.

    CarFinalDressSpec.jpg
    1000 x 773 - 330K
    SpecularRaw.jpg
    1000 x 773 - 56K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    And here's the specular pass after I applied a mask to eliminate the bright specular.

    Again, of course you can adjust the amount of gray in the specular image to just dull the specular a bit, or, like I did, make it almost black so there is virtually no specular.

    SpecularMask.jpg
    1000 x 773 - 57K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    And, for completeness, here's the actual mask that was applied to the specular pass...

    Mask.jpg
    951 x 739 - 26K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited April 2015

    Another thing you might want to try (although with this image, maybe not cuz the reflections are kinda cool, but anyway...) is to blur the reflections in the floor. Why? Just cuz. :) :)

    It's an interesting and useful technique to develop, since doing blurred reflections in a 3D render usually takes a lot of time.

    For this image I used the all important "object index" pass. Assuming most are using PS, you can first go into the image channels and make a duplicate Reflection layer. Then click on the object ID channel, and then use the selection tool with a low tolerance (maybe 0), and select the floor object. Once that's selected, jump back to the Layers and apply that selection as a mask to one of the Reflection layers. Then, invert the selection and apply it as a mask to the other Reflection layer.

    Why? Well, we're only going to blur the floor part, but not the other parts (ie, the car reflections, etc.). So we can blur the floor reflections separately, then add the two parts together again. Or, if you're really cool, since you have the floor and "other" reflections separately masked, you can modify the "other" reflections however you want. Maybe boost their luminance to make them brighter, or give them a slight blur, or modify them so they aren't as clean, and follow the slightly uneven contours of the car body....

    So to do a straight floor reflection blur, go to the masked floor reflection layer and blur it. Assuming both Reflection layers are set to Add, you should now have a blurred floor reflection, but the rest of the reflections are untouched.

    FloorBlur.jpg
    1000 x 773 - 307K
    Floor_Blur_Levels.JPG
    248 x 166 - 18K
    Post edited by JoeMamma2000 on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited April 2015

    And here I applied a Brightness to the masked floor reflections just to give you an idea of some of the things you can do. Too bright, but I just wanted to show you what you can do.

    FloorRef.jpg
    1000 x 773 - 332K
    Post edited by JoeMamma2000 on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Oh, and I forgot...here's the raw Reflections pass image

    Reflections.JPG
    937 x 725 - 44K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited April 2015

    And here I applied a texture to the floor reflections, using one of the PS filters on the masked floor reflections layer.

    Really, there is no limit to the different effects and results you can get. You just need to keep thinking about isolating each aspect, and use your artistic ability to come up with great looking images.

    FloorTexture.jpg
    1000 x 773 - 365K
    Post edited by JoeMamma2000 on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    And here's a very quick and easy modification...I merely used Object ID again to select ONLY the dress, and applied an HSV adjustment using the resulting mask to give the dress a different color, while keeping the rest of the image unchanged. And I also used a different PS filter on the floor. And all of this can be done in seconds, with real time feedback, rather than having to re-render and wait for results.

    BlueDress.jpg
    1000 x 773 - 362K
  • MiloMilo Posts: 511
    edited December 1969

    This is a great thread! Learning a ton, thank you Joe and everyone who has been joining in. I am going to have to go back through this in more detail again.

    Thank you
    Milo

  • 3DAGE3DAGE Posts: 3,311
    edited December 1969

    HI Joe :)

    When you're in Photoshop,. (as you probably know) You can select parts of different layers to create a larger / more complex selection area,. such as the Dress, and it's reflections on the Floor and on the vehicle bodywork and headlight,. (I'm asuming that the Red colour on the headlight is a reflection of the derss)
    It looks like the flowery part of the dress.

    Then you can create the new "Adjustment layer" to adjust the Hue/Saturation on the dress floor reflection etc. and it'll aply to that "compound selection"

    I don't have the render passes, so i just roughly created a selection and adjusted the reflections.with an adjustment layer.
    also on her right boob area, there was a little bit of the red dress colour,. I don't know if that's a flaw in the selection, or the render pass.
    It's easy to clone brush those sort of minor issues out

    Hope you don't mind me posting.

    colour_corrected.jpg
    1000 x 773 - 370K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    3DAGE said:
    Hope you don't mind me posting.

    As a matter of fact I love it !! Finally someone actively participating. It gets a bit tiring talking to myself here... :) :)

    Yeah, as much as I love everyone here, there's a limit to how much time and detail I'm willing to put in to the stuff I'm doing here. There are tons of detail issues with all of these images, since the point of the images is not production perfection but rather to convey basic concepts.

    So participation is a good thing, even if it is to point out errors and issues. I just wish sometimes instead of pointing out errors and issues there was more learning and "oh, I didn't know that". But hey, I keep hoping there's someone out there who's following this and doesn't already know it all. :) :)

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    By the way, as a challenge for some of the more experienced folks out there...

    There are still a few render passes in Carrara that I haven't covered yet. Maybe some of the experts here can add to this tutorial using their own scenes. I think Fragment Coverage still needs to be covered, as well as material diffuse and specular. And maybe some more discussion of what you can do with the world position.

    Anyone?

  • Box8068_31c338ee4bBox8068_31c338ee4b Posts: 280
    edited December 1969

    Sometimes I think that people that do this for years tend to get in a habit of describing what they do in a sort of shorthand. .

    Absolutely true.

    But on the other hand (and it's something I've never quite understood), is this...folks who have very little background in a subject can ask what, to those experienced people, is an extremely ambiguous and open ended question, but their peers with little background can somehow understand what they're asking and focus in on the answer right away.

    It's like "hey, I've got this thingy and I need to make it better, how do I do it?"....and guaranteed, someone will step up and say, "oh, yeah, I know what you mean. Just download this plugin and push this button and it will do exactly what you need".

    And the answer is generally "Wow, exactly what I need. Thanks." And I'm generally sitting there scratching my head thinking "huh? Is that what he was talking about??"

    I *thought* I had described in great length with great clarity some of the ways in which you can do this stuff, but as usual I seem to miss the mark. I think it's usually the step by step detail with pictures that somehow is what people want or need. I keep thinking people can take general concepts and convert them to that detail on their own.

    Frustrating...especially after spending so much of my time trying to provide good information. Oh well, such is life.


    Joe
    I love this thread! Thank you. I do basic animation in Carrara to explain technical audio subjects. My renders are clear, but I do not strive for realism or beauty so I have not experimented with multipass renders or compositing. You explanations have been clear and I will do more research on these topics. Excellent!
    As a side note some of these render passes where the greyscale represents something other than light (distance) are just beautiful.
    8068

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    8068 said:
    As a side note some of these render passes where the greyscale represents something other than light (distance) are just beautiful.
    8068

    Yeah, I know....And I especially like this normals pass :) :)

    Normal.jpg
    1000 x 773 - 285K
  • evilproducerevilproducer Posts: 9,031
    edited December 1969

    I for one have a video project coming up where I will be doing a sequence of live action background plates and compositing elements from Carrara into them. I intend to use a few of the render passes that have been discussed. I'm not sure which ones at the moment, as I'm still in the planning phase, but I will keep you all posted.

  • evilproducerevilproducer Posts: 9,031
    edited December 1969

    8068 said:
    As a side note some of these render passes where the greyscale represents something other than light (distance) are just beautiful.
    8068

    Yeah, I know....And I especially like this normals pass :) :)

    I wonder if the Normal pass was broken in C7.2 like the Shadow pass was? I tried a Normal pass once, and it was very noisey, almost like banded static. It looked nothing at all like your pictures, Joe. I'll have to try that one in C8.5 as well.

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Okay, well assuming that nobody is bold enough to step forward and and at least try to add to the unfinished parts of this tutorial, let me get my spoon out again.. :) :)

    There is a render pass in Carrara called "Fragment Coverage".

    I can hear everyone saying "Huh??"

    Well, a "fragment" is another word for "pixel" which is another word for "picture element". Which is the individual square elements that make up your image. And each pixel can only be one color. And if you zoom into one of your rendered images you can see "jaggies", or jagged edges where the renderer has to decide what color to make each pixel. And when there are sharp edges of, say, an object against a background, it's tough to make that a clean edge when your image is broken up into squares. So what you can do to smooth some of that is called "antialiasing", or other mathematical magic to make things at least appear like a clean, sharp edge.

    So what is "pixel/fragment coverage"? Well, if you render out a "Fragment Coverage" pass like the one below you can start to understand what it is. In the image below it looks a lot like the "find edges" feature I mentioned previously in this thread. And I mentioned that you can use it to generate a thin outline mask around an object so you can do your magic to blend that object into the background.

    Fragment coverage basically gives you a grayscale image that is the output of what the renderer uses when it is antialiasing the edges of an object where it overlays another. If you look closely, it is a gradient outline on a white background.

    Is it useful? Well, sure. If you combine an Object ID pass, which describes the object itself, and the coverage pass, which describes the edges, you can do cool stuff like feathering or darkening or lightening or smoothing edges. Will anyone here use it? Probably not (like many of these passes), but now you know what it is at least.

    I've also included the coverage pass overlaid (semi opaque) on the final render image so you can see how the two align.

    CoverageOpaque.jpg
    1000 x 773 - 273K
    Coverage.jpg
    1000 x 773 - 93K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Oh, and for the next episode of this tutorial, it will be a team presentation of the Material Diffuse and Specular passes, one of the easier passes to understand.

    And the team members for this presentation will be Evil, 3DAge (in a comeback appearance), PhilW, Headwax, Dartanbeck, Wendy, and Rashad.

    Ladies and Gentlemen, put your hands together for the Dream Team !!!! :cheese: :cheese: :cheese: :cheese:

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Oh, and here's a zoom of the fragment coverage pass showing jaggies and the gradient nature of the outline.

    FragmentZoom.JPG
    925 x 767 - 58K
  • DiomedeDiomede Posts: 13,288
    edited December 1969

    Thanks, Joe. Very interesting. Please keep it up.

    For others like myself, there are numerous questions that are really about how to do this in another program. Naturally, there are resources for compositing in other software on forums dedicated to other software (no surprise there). Here are some tutorials and similar videos that I think are related. Hope it helps others like myself.

    Compositing concepts and guidelines in Photoshop - based on stock images but concepts are general (hour long video)
    https://www.youtube.com/watch?v=38Ku7XA3LKI

    More specific example based on depth of field render pass (Photoshop)
    https://www.youtube.com/watch?v=kfrurfMqbJU

    Another Photoshop compositing tutorial - more focus on Photoshop tools (23 minutes)
    https://www.youtube.com/watch?v=szQ0kj1xuZc

    Something focusing on multipass rendering (Vray to Photoshop, but more geenral) - demonstrate using light pass but more general
    https://www.youtube.com/watch?v=q0WpIzi8sc4

    High speed, a little off topic. I just thought this was cool. Shows workflow of modeling (Maya), texturing (Photoshop and Mari), rendering (Mental Ray), and compositing (Nuke).
    https://www.youtube.com/watch?v=fhVzcYbyETY

  • evilproducerevilproducer Posts: 9,031
    edited December 1969

    Oh, and for the next episode of this tutorial, it will be a team presentation of the Material Diffuse and Specular passes, one of the easier passes to understand.

    And the team members for this presentation will be Evil, 3DAge (in a comeback appearance), PhilW, Headwax, Dartanbeck, Wendy, and Rashad.

    Ladies and Gentlemen, put your hands together for the Dream Team !!!! :cheese: :cheese: :cheese: :cheese:

    As I said above, I have plans to use render passes in an upcoming video project. When I get my background plates and other assets together, then we'll see what I have time to do. In the meantime, keep your sarcastic comments to yourself.

Sign In or Register to comment.