[simmsimaging]

Hey All -

I've seen quite a few threads lately that all touch on a common and important subject - photo-realisim - but in particular how it affects aesthetic choices for renders. I'm pretty new to the MW scene, and working directly in 3D, but I see a lot of people trying really hard to make "photo-real" images, but doing so with what seems to be a pretty narrow view of what counts as "real". I think this is a big issue most particularly for MW users because we all get so caught up in the amazing *realism* this renderer can produce, but not enough of us are producing amazing *images*. Don't get me wrong though: there are many, many exceptional images out there and all of you have produced some amazing stuff.

This is not a failing on the part of all of us working with MW, or with MW itself either - its just that after playing with MW for a while many of us learn something that pro photographers already know: Maxwell may be as easy as taking a picture (yes, arguable ), but taking a *good* picture is not really easy! Often not even remotely easy.

I thought it might be good to start a thread to discuss the issue and maybe we could all share our POV's and tips about what it takes to make not just photo-realistic renders, but *good* photo realistic renders. I think it's really the same stuff that makes good photographs - and paintings/illustrations too for that matter. Obviously there will be no carved-in-stone rules because it's different for every image, but general principles do apply and rules of thumb are always handy to have - even if you choose not to use them.

Anyway, with all that in mind I'll kick it off with a couple of posts and see what comes back

[NicoR44]

Hi Brett,

great idea, I will follow it with great interest!! and maybe also participate as well

[simmsimaging]

A very important thing to be aware of, as I see it, is that you don't have to separate "visually striking" from "photo-realistic" to have great renders/images that also read as realistic. Extreme devotion to photo-realism without a good broad knowledge of what that can mean will probably only hold you back.

Don't get me wrong: it's unbelievably cool to see renders that are MW realistic, and we all need photo-realism (or we wouldn't use MW), but once you get over that initial fact there has to be something more to the image. Most of our clients are not going to be happy with what amounts to a highly realistic version of a somewhat mediocre photo Put it this way: your render may pass as a photo - but would that photo get you work as a photographer? If not, then maybe your renders could use some more work. Depends on your goals and time I suppose, but rarely do people complain that images are too nice

I think a good step is just keeping in mind that realism is only one goal with our renders, and we don't want to *replace* that with another, but we may want to expand what that actually means.

Educating your eye to be able to *see* the difference between average and good, or good and great images, so you know how to improve your own work is not easy either. This means going beyond looking at photos to see what makes them "real" and looking to see what elevates the better ones.

One really easy way to start getting a feel for photo quality is to pick a subject and do a search on istock (http://www.istockphoto.com) Then do a search for the same thing on Getty under *rights-managed shots only* (http://creative.gettyimages.com/source/ ... mages.aspx ) Be sure to tick off 'rights managed only' so that you
see only their best stuff. Compare the overall look and feel of the shots and see what you think. You should see a pretty pronounced difference over a number of shots (although istock has gotten MUCH better these days).

It's a good way to start anyway

[simmsimaging]

Great NicoR44 - looking fwd to your input!

b

[Boris Ulzibat]

I guess achieving this "balance" between 3d photorealism and "good picture" must begin with learning something about photography itself.
And i do not mean DoF, f-Stop, Aperture values, focal distalces etc.
I mean rules of composition, colours, human perception of shapes, textures forms, etc.
Then, first you build a scene looking as a 3D artist/architect, etc.b but when it comes to rendering, you must look at your scene as a photographer.
Your 3D app's viewport is a camera viewfinder now.
Just some thoughts... I am not a good photographer yet, but i'm learning

[Jeff Tamagini]

Great topic indeed Brett

I have the constant struggle at my office when I do renderings of my principals wanting me to cram say, more of the room into the rendered view. Then when they see it they always complain that it looks all distorted and they do not want to take the responsibility that it is their request that for more that I needed to use a 15mm "lens" on my rendering. The photographic composition just become lost, but they are starting to finally come around and trust me.


Just remember its all about the rule of 3rd's

[simmsimaging]

Boris - That's exactly the kind of things I mean. Learning to become a photographer is tough but a necessary part of creating really great images on a consistent basis.

Jeff: Rule of thirds is a great thing to keep in mind. It's tough to figure out where to crop the image and what to include. It can sometimes be tempting to try and show everything in there, but focusing on key areas of interest is often better. Depends on the situation of course.

Have a look at Tora's site (purerender.com) and look at the slideshow of shots. If you look at what shows up in the slideshow versus what you see when you open the full shot up you'll get a sense of really good cropping for visual impact. There's nothing wrong with ANY of those images, but some of them are actually much stronger visuals cropped in and composed for the slideshow.

Just my opinion, but if they didn't agree they probably wouldn't have cropped them that way

b

[simmsimaging]

Another area to look at: colour and contrast levels in our images.

A great many MW renders seem to suffer from too low contrast in terms of both value *and* colour. So many that I see look dull and flat - like they were shot on a cloudy day with old Ektrachrome 100 on a so-so 35mm camera. Real enough, but only one way of seeing real, and not the most exciting.

Not every image has to be flawed to look real, and not every image has to be "documentary" style to be real, and there's such a range of what that even implies: (Compare some newspaper photographs or real-estate shots or typical catalog shots with high end photography. Look at National Geographic, a high-end advertising shot, etc. There's a wide range of what "photo-real" can entail)

I would say that to improve the visual interest of your render do not be afraid to experiment with increased contrast and saturation. Photographers do it all the time. It helps to create a visually interesting shot regardless of subject type (have a look at Geret's work for just one set of examples).

Use the full range of values in your shots wherever possible (good highlight values, good shadow values). Rich tones are not just "fake" - they are often accurate to a subject and reflect well controlled exposure and lighting. You can also use saturation selectively to control areas of interest in your shots.

For sure, there are certainly beautiful photos and renders done with limited palettes and limited contrast, but they are exceedingly tough to do well, and often represent a very particular style. If that's not your style, you probably just need more contrast in your image :)

Things to do/watch for:

1) In your renders watch out for cyan/blue casts which often (I think) come from the physical sky or even just really neutral emitters (again, I think - don't have enough flight time in MW to be sure yet). That cast *can* look very "neutral", which some want, but that particular cyan/blue shift I see in many renders tends to really dull out brighter tones and gives everything a more dull/dead look. Just going into Pshop and dialling back the cyan a few points can really help, but slightly warming up the light helps a lot too.

2) Avoid over-filling your lighting - a shot needs good 3/4 and shadow tones to have tooth. If you have to increase your "ambient" or fill light in order to clean up your renders faster then go back in Pshop and deepen your 3/4 and shadow tones a bit to compensate, but you still have to have the *quality* of the light there or no curve work will bring the right contrast back. Lighting is key. On the same note; while it's good to have shadow detail, you have to ask yourself what is really important in there? How much shadow detail do you need? Is a hint enough? Often you just have to have detail in one area, the viewers mind will fill in where it's repeated in the image. Just make sure your shots make use of the full value range wherever you can.

3) Try to learn the difference between colour contrast and value contrast. Good images use both, and having control over both will give you a lot more options. Most attempts to fix contrast in post don't keep this distinction and you end up with shots that are overly contrasty and/or overly saturated. Use curves in Photoshop rather than levels or Brightness/contrast, and try the Fade command to Luminosity mode only, or Colour only to see what difference it makes. In any event, while it's better to get it in your lighting, it's not always possible or easy that way, so don't be afraid of tweaking in post!

[Thomas An.]

Very good topic and thank you for all the useful hints Brett !

There is indeed a difference between "photoreal" and "aesthetically useful".

Reaching the photoreal barrier is only the beginning. Once there, then it is all about composition. One is then judged as a photographer and there is a large quality gap between home-made photos and world-class images.

[mverta]

Holy Christ, the last thing people need to be doing is crunching levels for contrast and boosting saturation in Photoshop. You just defined the formula for why 90% of CG images look like shit.

To this day, I will still film practical models in a studio over CG approximations 7 days a week and twice on Sunday if I can, because you get a billion things for free in camera, and post-ing a real photograph is a completely different story than posting a render.

As much as we liken Maxwell to being a virtual camera - and it is surely the closest yet - treating its renders like photographs is a gross overestimation. The community as a whole needs to learn the relatively new skill of seeing work/creating work through photographic sensibility, absolutely. The skill sets of lighting and composition and camera work have suddenly become forefront and relevant; this is a beautiful thing.

And surely in time, renders will become nearly 1:1 with photos such that universal post-work techniques apply.

Until then, levels adjusts, saturation adjusts, balance adjusts, highlight adjusts - all of it - are the compensatory cheats of an antiquated approach, and source of a billion bad habits.

Post is too powerful a tool to be placed in the hands of someone who doesn't know how to get the look without it. My recommendation for a growing community with a growing sensitivity is: have the discipline to create the best possible image you can, in render, before you ever go to post. Post will always be there - and you'll be able to make more effective use of it when it is truly accenting material that absolutely works on its own.

You have to learn to crawl before you can walk.

_Mike

[andrebaros]

Great thread.

Personally, I think that the key is in the story telling. Does the image draw you in. I was looking at some work by Gregory Colbert over the weekend and noted several times how flat his images were but how the narrative, composition, and framing blew my mind away.

Here's a link to the images... and even though it's technically moving, he holds the frames for so long that you could call it a series of stills.

http://www.ted.com/index.php/talks/view/id/15

All told, I'm also a slave to the rule of thirds, and because I only shoot architecture, also to trying to balance foreground, middle ground, and back ground.

[Tora_2097]

Hi Brett,

That's a great topic indeed, I could ponder about it all day. Thanks a lot by the way for the compliments on our company's website, glad you like it.
As we are a company that has of course to make its profit, the site's content has to be focused on the typ of client you want to attract plus a bit of showcase stuff (the cool bits )
I often face a dillema when working for clients, similiar to what Jeff described above. The clients we deal with have a rather limited sense for visual impact -or better said they have a different focus of mind when looking at images we produce for them. They are less interested in appealing compositions, balance between shaded and lit areas etc. Instead they want to deliberately show everything possible and make all areas bright. A 15mm lens is not an uncommen thing to use here at work (11mm is my record so far...)
When I am doing works without a clients pressure I usually decide wheter it is a true photoreal image (the drawer image I made for example) or more of a visually strinking image (my kitchen nightshot perhaps).
There are q few things to consider when you are after photrealism and I found them quite accurately listed just recently by Dru Blair:
Element of photorealism:

proportion,
color fidelity,
scale,
perspective,
edge transition,
subtlety,
non-contrivance (chaos),
object relationship,
light and shadow,
reflection,
radiosity,
transparency,
texture,
atmosphere,
depth of field,
weathering (wear and tear),
flaws,
detail,
visual sense
Thats quite a bunch to consider. Another thing to consider is the complexity of your scene. It is much, much easier to make your productshot photoreal as a fully cramped 400m² apartment with exterior view. As the scene grows you'll naturally introduce more and more variables that could potentially break your photorealism.
If I decided to create a photreal real, it is a key component to gather reference material of the thing you are after. I cannot emphase this part enough, it is truly crucial. I model the thing then quickly with a few boxes to determine :
a) focus
b) masses
c) light and shadow

The next stage is to replace those proxyopbjects with the actual models and there are few things to consider as well:

a) NO sharp corners, (bevel,chamfer everywhere)
b) randomise
c) chaos

Texturing should be considered at least as important as modeling. You can make a poor model shine with great textures, but even a great model can easily be ruined by a bad texture. It is absolutely required to draw your textures with photoreferences the versatility created by mother nature is hard to mimic otherwise, just think of wood or stone for example.
And lastly I would encourage everyone to use as much postprocessing as necessary. A renderengine can only take you that far- no matter how "physical correct" it is. A photographer surely does not decline of altering their images in photoshop, so why should we?
You'd be surprised -or shocked - of how much in many of my images is actually altered or even completely exchanged later in photoshop or other softwares. The drawer image of mine has literally no surface that hasn't been manipulated in photoshop. The list goes from changes in hue/ saturation to complete retexturing of some faces. Metalscratches, rust, dents, dirt etc. can also be easily added in photshop. It took me quite a while to let loose of the concept of doing everything in 3D when it could be done much faster and with much more control in photoshop.

Kind regards,

Benjamin

[Tora_2097]

@Mike: You posted while I was writing and it seems we disagree on some concepts here.
It is of course wise to start from a basis as good as possible, but with experience there comes a point when you know what can or cannot be achieved as a posteffect and I stop refining the scene exactly when I feel that I loose more time with rendering than gaining benefit from that.
And that is when I move over to the postprocessing stage, knowing that the desired affect can more easily/ faster achieved in Photoshop.

[tom]

It's like the story of "blue pill or red pill". It's a matter of pushing the limits before or after the post. I hear and respect Brett here in terms of required knowledge but I strongly agree Mike in terms of choice. If possible, we should avoid post modifying a render to turn it into a photo. Oh, but if we can make a photo-like render before using all the tricks, then any post fx is welcome on that photo to make it look attractive and more convincing.

And my key points -before hitting the render button- are:
1) Modeling must be accurate and imperfect. Everything, even a filleted edge should be slightly non-uniform. Sharp edges already not acceptable even for an ipod shuffle model. Viewing distance won't help hiding these facts, they must be done.
2) Dimensions must be correct and camera facts must be used in terms of lens and film parameters.
3) Observer location and camera orientation must be natural.
4) Materials (including textures) must be perfectly calibrated no matter how much time you spend on them. They directly affect the look and a small incorrect value may turn it to something completely unreal.
5) You should avoid attempting to materialize/arrange/light and render the full scene at once. You need to do several steps with simplified setups for calibrating each.

and so on.

[simmsimaging]

Hey guys -

Tom: Great input, thanks. I think you may have hit on the point where Mike and I diverge: This isn't about using post to make a render more *realistic* (although it's certainly possible) what we're talking about is your second concept: how to make photo-like renders *better* - however you do it.

Andrebaros - great point. The Ashes and Snow work is amazing - look around for the website and see the actual print work - unreal. Literally in some cases, but beautiful stuff. His composition is phenomenal.

Benjamin - great post, thanks for sharing your thoughts and insights. You are right, the subject is vast - limitless really.

Mike - Thanks for your input as well. I'm with Benjamin though - firmly in the other camp when it comes to Photoshop and post work. I disagree with the blanket statement about 90% of renders being shit, and that they are because of too much contrast/saturation. That sure doesn't hold up with Maxwell images at least - most aren't so bad really, and those that are problematic swing the other way. Whatever the proportion actually is, a good chunk of the ones that are in fact bad are so because of lighting/composition issues. Contrast and saturation are only part of that picture.

I have worked in "post" for good long while now and in my experience thinking of Pshop type adjustments as "compensatory cheats" etc is the antiquated view. It's a tool - use it where it fits best, and that changes all the time. As Benjamin said - some stuff is just faster that way, and some stuff is just better or more flexible that way. You clearly feel differently, but I work with top notch photographers all the time - and quite often we start planning the "post" before shooting even starts. It's not because we don't know what were doing Photos often need post - how renders, no matter how good, can have a blanket exemption is beyond me.

That said, I will agree with the essential point that using post techniques badly is not a good thing, but rather than throwing it away or putting it in a very small box, I would suggest learning to get better at it - which really means improving your eye. If you can't do a good job with post techniques chances are high you won't do any better in your renders with lighting and materials because you don't have the eye for it yet no matter where you try and build it.


That really just restates the problem this thread is hopefully here to address though. The thread, I hope, is not about whether to use post or not, but rather the type of things it takes to make a better image whichever way you go.

b