Adding People to Renderings

[xEndlessxUrbiax]

I'm wondering what you guys usually use to add people to renderings. In the past I have always added them in post (photoshop) which usually looked fine because my the people in my scenes were far enough away that the detail didn't need to be there. I am asked by my firm to do more and more interior renderings in which I find it easier to just omit people out of convenience. Every time I try to add some in post they look like they don't fit in so I'm avoiding using them altogether. I've read a little about RCPs and downloaded a few but I haven't been able to import them into my 3dsMax scene. From doing a little research it looks like you need a licensed plugin to do this. Another option I have tried with a less than decent result is finding a 3D modeled person on the Google Warehouse and rendering them out in my scene.

 

How do you solve this problem? I'm willing to spend a little money if it is needed to download a set of 3D people for MAX but it seems like that might be overkill.

[Umesh Raut]

If you are prepared to spend, then that RPC plugin might be of some help with their free stuff, but the other and more pocket friendly way is to photograph some people in similar surroundings and use those snaps with their alphas in max scene as mapped on to planes facing the camera. That way you not only get them blended as well as lit with your scene lighting properly but are also able to scale the planes where necessary.

[Boris Horosavin]

I too hava problem when inserting 2D people in post, they always turn out crappy, no matter what. And, also, I decided not to use them, unless specifically asked. As for 3D people there are some relatively cheap low-poly models out there, Evermotion and Dosch had some if I remember correctly... They are very simple and light and won't overweight your scene, but from close they don't seem very realistic and are suitable mainly for background. As suggested, you can use pics of people and alpha maps for opacity, to "cut them out", orient them towards camera and put them directly in scene for rendering. However the shadows they cast are not accurate and overall effect was not satisfying for me- actually it was very similar to adding them in post processing via Photoshop.

[postite]

For the "planes facing the camera", I usually use the animation constraint "LookAt Constraint".

 

Just my two cents

[Boris Horosavin]

Cool man, I remember using that some time a go, but then I totally forgot what it was called and where it was (mainly because I rarely use animation tools). Thanks.

[LukeC]

Generally I find it takes a bit of time to find people with the correct lighting to sit right in your scene in post. You may find it possible to search stock photography websites for a similar scene or lighting scenario that matches your 3d scene. Additional to this I find it helps to put 3d people in the scene just as shadow casting objects so your photoshop person sits better in the scene.

[Kyle_Grace]

I think Luke is right regarding this. 3D people always look awful no matter what the polycount, personally I think RPC's always look ropey also. Rendering in a 3D model that gives you the correct lighting/shadows is a great tip, then add your person/people in post.

 

Adding in post is the best way, but in order to do this well you need a decent library based upon lighting conditions. Often stock libraries of people who are 'posing' tend to seem a little cheesy.

 

Have you tried the following:

1 - Select person based on how they are lit (ambient, shadow, Interior etc)

2 - Settle them into the scene either by using Luke's tip of rendering something out for GI / Shadow purposes or faking it by painting it up.

3 - Blur the very edges of the person = CTRL + Click the LAYER to select the outline, then go to Select > Border and select something between 3 to 6 pixels depending on how big they are in the scene, and gaussian blur the edges. It's subtle but it stops them looking like a 'cut out form a magazine'

4 - Take a copy of the layer and add it on top as 'Multiply' layer type, and use this to control the intensity. This often has a knock on effect on saturation, to stop this -desaturate the layer on top.

 

If this is something that people are interested in I might do a tutorial as it never really seems to be covered.

 

Hope that helps.

[xEndlessxUrbiax]

This helps, thanks guys. I'm sure this thread can go a long way in answering this question as it is a common one in this industry.

 

It seems like the preferred method is to add in post, which is how i have done it in the past but with these current renderings, I couldn't find the right person in photographs found on the internet to match my scene so I opted out of using them. I'll post my rendering so you can see what I'm talking about.

 

Thanks for the replies, I'm gonna try taking the pictures myself from now on, maybe using some makeshift green screen or something.

[Crazy Homeless Guy]

Sitting people are the most difficult things to composite. You will definitely need to shoot custom people for the views you posted.

 

At this point I would not bother with a green screen, you have more difficult tasks to work through. I would start by simply pulling out your camera and shooting test shots of a subject to see how well you can even replicate the camera in the scene with the physical camera you will be using.

 

Once you are comfortable with the angles you need to shoot, then start rounding up friends to pose in this position. Ideally you will want them sitting on a low stool while keeping their posture relaxed enough to make it look like they are sitting in a chair. You will also need them to position here hands so that it looks like they are working on the keyboards.

 

Why a stool? you want to keep nearly every part of the person visible to he camera lens. Trust me, you do not want to paint in parts of the body that were blocked by armrests and backs of chairs.

 

Basically you have choosen a difficult shot to learn how to composite people.

[xEndlessxUrbiax]

Thanks Crazy Homeless Guy, the stool is a good idea, I probably only would have thought of doing that once I started doing selection deletion of the chair in photoshop and ripping my hair out.

[steyin]

You could do a black and white renderpass with 3d people (for scale and position reference) to overlay in Photoshop or whatnot, then add people in post who match up best. Rami has a process I tend to follow for people:

 

http://www.ramyhanna.com/2010/11/3d-process-post-process-tricks-revealed.html

[Fooch]

I'll add my 2cents + tips

Here are a few major flaws of cutout ppl in any architectural scene and how i would solve them in post:

 

Dreaded halo.

--> crtl click the layer, use refine edge to delete the edges .. and mask that selection. Here's the important bit : Paint back the parts which is obviously missing (it might take away wayy too much pixels) and only then , apply that mask.

--> If its a back layered person (someone far away), I can speed it up by feather the edges --> crtl click, shift crtl i (invert selection) feather (shift f6, or crtl alt D for older versions of photoshop --> seriously why did they change it, feather it by say 1 or 2 pixels .. then delete. if its still halo-ing, delete again. I only do this for smaller people on the back or mid ground.

 

Lighting , lighting lighting + Placing ....

--> the common mistake for people using stock images is the lack of lighting or people being weirdly placed. Here's one tip to help the work flow.

I divide a scene to Foreground , Mid ground and background Ppl. There might be more if needed.

I start by composing the scene from the foreground (stupid to make a busy background when a foreground character would block it)

The FG ppl should relate to each other.. etc, ppl should be in groups conversing .. lighting must sort of match (can be fixed) .. plus no dodgy res stuff. Plus less is more should apply. Having a few good fg ppl will really tell the story of the space. A good tip to learn is to use cinema as a language (IE. keep things in the 1/3rd rule, frame your shots, compose ppl correctly instead of weird putting personal around)

These all go into a group called FG. Apply this for Mid ground and background (all in separate groups)

 

Lighting.

I would start with the FG group first. This is important. Set the group blend mode to normal.

Any adjustment layer within the group would only apply to the objects within the group. This allows heaps of freedom to light the ppl.

I would normally make a shadow adjustment curve (pulling the rgb down, and perhaps tweaking it to push it to a certain colour) and invert the mask from white to black. Ditto for a highlight adjustment curve. Make sure these are on the top of the group.

I might make an overall colour match adjustment too, perhaps a hue sat or a curve.

 

Now here's the simple bit, use the brush tool , select the mask, press D which sets it just to black and white , set it to say 30% (press 3, this will set it) and with maximum softness (lean it to the left).. and Paint. The inverted / black mask is invisible .. so by painting on the High light and shadow curves, you can create and match nearly any lighting situation on even tone cutouts.

 

I would focus on the fg ppl. The mid ground and background ppl would have the same treatment, but it will be done quicker. Example, to tone and match each fg person, I would use adjustment layers on the main ones and link it to them (crtl alt g, links the top layer to the btm) But the smaller background ppl would get an overall tweak, or direct adjustment (without using adjustment layers, which i tend to use them alot because I prefer to keep my workflow open and tweakable)

 

Shadows

Here's one important thing about shadows. Shadows dont multiply on other shadows.

Test it yourself. If u take 2 objects the middle bit of your shadow does not become darker.

 

Now this is important because most people do shadows this way. Transforming a copy of the same layer thats changed to black / blurred and reducing the opacity and setting it to a blend mode of multiply.

I would still use that however, just becareful. If that shadow goes over another shadow, you get a strange spot on screen which has darker patch in the middle of it. This doesnt happen in real life.

 

Background ppl --> I would use an adjustment shadow layer for those and paint the shadow in. You can change your brush shape to a thinner skinny line. This makes it easy to paint it on an adjustment layer for the BG ppl. plus the benifit is with this, shadows dont multiply.

 

FG ppl --> I'm a little more careful with these. Its a mixture btwn duplicating the cutout layer , adjusting and fixing it. The important thing is getting it to match the feet location. I do paint and brush in parts of it. Plus masking the edge of it off. If its an interior or a night scene, take note .. interior lights cast multiple shadows , say 3x shadows for a night -ish scene.

 

Grounding Ppl

Here's a tip from the film industry that is missed by a lot of arch viz teams. Ground your ppl.

If you take a photo , zoom into the pixels and you would notice that feet is grounded to the ground. There would be a darker layer of shadow under the feet ... just a touch darker.

This helps the floating ppl problem in arch viz. Just make an overall adjustment layer below all the ppl groups, get a small thin brush with high softness and paint in a darker adjustment for all the feet. A small touch is needed to ground everyone in the scene. Its great for sitting ppl as well.

 

Movement

Post Motion blur is good. However it looks really weird if its done on the whole person. Seriously, no one looks like that in a photograph.

The trick is to recreate an open shutter image. I would duplicate the cutout layer. do rough selections on say the moving part, like the hands or front of the legs (only on the fg ppl), feather and radial blur / motion blur those parts. Then I would mask it out (Make a mask on that layer and convert the mask to black , crtl i to invert) and paint in where i want my blurs to be.

 

 

Tips

Lastly --> saturation. A huge flaw is to use heaps of photographed people which one or 2 might wear a really sharp colour.

Do the squint test where if you step back from the monitor and squint your eyes. If something colour wise pops up which isnt intentional (sometimes i do want sharp colour focusing on certain parts of the scene), selectively desaturate that layer (with an adjustment layer , so you can mask it back)

 

 

 

Hope it makes sense and it helps

[Fooch]

Plus of course, scale...

 

If you take a photo at eye level , with the horizon right ahead , most heads will be at the same level.

I would always get a low res , quick and dirty height render (cylinders in max scattered around just to give me a grid to work with)

[xEndlessxUrbiax]

Thanks Fooch, definitely more than two cents

Awesome advice and well organized. I'm gonna keep referring back to this thread whenever I get back on doing renderings.

 

Good stuff

[Fooch]

Oh another thing,

Use vray multimatte elements / rendermasks to generate masks so you wouldnt have to cut out say ppl behind stuff.

You can always put groups (example = masked with a render matte) in groups (painted on mask) ..

 

Tip for mask selection from vray multimatte elements , quickly selecting the RGB channels.

CRTL ALT 1 , 2 or 3 will do a selection of 1 = red , 2 = green, 3 = blue (my colleague laurence showed me this a few months ago and its brilliant for saving time)

[Thox]

Very nice tips Fooch, some of these I already do. However, I always have problems with people behind glass, they never look convincing, any tips for this?

[Fooch]

thanks nick,

 

What i normally do is this.

IF there is ppl behind glass within the "group" of ppl, (set to normal for the blend mode) , I make an adjustment layer / curve or hue sat to lighten / tint and desat. (this can work in a group too). Use the glass matte (from vray matte elements) to mask it out. Sometimes you might have to cut certain bits out (example ballustrade glass over a glass shop can be a pain, you might use rendermask from scriptspot to create a specific mask).

 

Now because its a masked adjustment layer / group .. this means we can move ppl around easily behind it. This is great for ppl behind ballustrades etc.

 

For really tiny bg ppl, I merge the bg and motion/directional blur it as well. Plus i might mask some parts out to to give a hint of movement behind glass.

[Boris Horosavin]

Well, you people obviously know a lot on this topic... However, I must confess that I never really understood this obsession with adding cutout people to 3D scenes that is common for most of architects on all continents. For an architect, which nominally I am not, Photoshop added cutout figures are a must-have, something without the scene is just shallow and incomplete.

For me main problem is the fact that no matter how well applied these people always feel like phonies. Even when a scene is absolutely photo-realistic and figures well inserted, adapted and organized - Fooch and rest of you guys obviously know the stuff - there seems to be a sort of contextual displacement that I simply don't like. I guess its because every 3D scene, no matter how realistic, is always at some level stylish and "synthetic" and these cutout models usually look too casual, inappropriate and chaotic - like a dadaist collage.

Therefore, I find that method I like best is this: Use alpha mask to cutout silhouette directly into 3D scene and then fill it with some abstract simple texture, like stripes or squares or whatever; some transparency or opacity added there helps these models fuse with the rest of scene. This way I get a feel of human presence in the scene without ruining overall equilibrium and stylish effect I like. Someone already noticed in a comment that Evermotion used this method quite a lot (I just don't remember who it was, since this post has been going on for days...)

Edited March 11, 2012 by Horosavin

[marius e]

I cant agree, I have seen some stuff thats great. A rendering will never be perfect, its a computer image, not real world. Not everybody wants it to be realistic. Its nice to have something that in between. Its also nice to have something which one can see is cg work but also looks real in a sense. Anyway, 2d ps people look better than 3d people by FAR.

 

Not just about cutout people, but about time. If the client doesnt like some people you can just delete the layer, with 3d people you have to rerender. Also have to render to see if the people poses are good time it takes to place them etc etc.

 

Some get it right others sort a, and some dont I suppose.

 

Well, you people obviously know a lot on this topic... However, I must confess that I never really understood this obsession with adding cutout people to 3D scenes that is common for most of architects on all continents. For an architect, which nominally I am not, Photoshop added cutout figures are a must-have, something without the scene is just shallow and incomplete.

For me main problem is the fact that no matter how well applied these people always feel like phonies. Even when a scene is absolutely photo-realistic and figures well inserted, adapted and organized - Fooch and rest of you guys obviously know the stuff - there seems to be a sort of contextual displacement that I simply don't like. I guess its because every 3D scene, no matter how realistic, is always at some level stylish and "synthetic" and these cutout models usually look too casual, inappropriate and chaotic - like a dadaist collage.

Therefore, I find that method I like best is this: Use alpha mask to cutout silhouette directly into 3D scene and then fill it with some abstract simple texture, like stripes or squares or whatever; some transparency or opacity added there helps these models fuse with the rest of scene. This way I get a feel of human presence in the scene without ruining overall equilibrium and stylish effect I like. Someone already noticed in a comment that Evermotion used this method quite a lot (I just don't remember who it was, since this post has been going on for days...)

[Boris Horosavin]

Anyway, 2d ps people look better than 3d people by FAR.

With this I do agree.

[Fooch]

With this I do agree.

 

The thing that grates me most is wooden axyz ppl in stills .. grrrr....