Archi-Tech - On-going Project

[Tim Nelson]

WOW, that's quite a huge difference. How are you going to do the animation?

 

That's the only negative about so much post work. It's fine for a still, but when it comes time to do animation you run into a whole different animal.

[Alex York]

Cheers Enrique

 

I usually have a rough idea in my head of what I'm trying to achieve, yes. But it doesn't go much further than that as far as colour and mood go.

 

But when it comes to detail, yes, I'm usually constantly thinking "can I get away with doing this in post?" while I'm working. The volumetric light is a good example of that.

 

Also, I'm not too bothered if there are artifacts and strange areas in the render. As long as they are not too large or complex then I will leave them for post.

 

Tim: yep. you're right on the nose with that. The only problem with this approach is that it would be useless for animation/image sequences. The only real thing you can get away with doing in post for animation is colour-correction and effects. I spent a good 5 hours painting out errors and altering the size of some of the objects, in post. I couldn't have done that if this were an animation sequence.

 

I've never done archi-viz for animation, so I guess I'll cross that bridge when I come to it

 

I'm not going to animate this any more. It just wouldn't be feasible. Not because of the post problem (I'd just need to spend more time making everything perfect before I render) but simply because rendertimes are immense. It took around 11 hours to render that image @ 1k. Not to mention that I didn't build this school as one single mesh. It's a whole bunch of individual scene files for each shot. I do have a finished scene with all of them together, but the rendertimes would just be too much. The deadline is 1 month including the report.

 

Shame, as I'd have loved to animate this.

 

Thanks again

[evanmoses]

Hi Alex,

 

I was wondering if you could maybe share some info about your photoshop workflow or pretty much anything else about the post-processing work you do, given how ridiculously good it is.

[Ernest Burden III]

That's the only negative about so much post work. It's fine for a still, but when it comes time to do animation you run into a whole different animal.

 

Not necessarily. What do you see as new problems with post-driven images used for animation?

[Alex York]

Ernest: the kind of post I mean (and I guess Tim means too) is paint work. ie. everything other than colour-correction and image-wide effects (such as fog) which are easy to do in post (and probably should be done in post).

 

but when it comes to paint jobs... it's no good for sequences, since you have to use rotoscoping and tracking to make your paint job work across multiple frames. example: you have a window, the bottom right corner is badly aliased over 10 frames. you could paint it, but you'd have to either hand-animate it or track the paint object and reshape it perfectly every frame.

 

post for this sort of work is limited to colour-correction, large effects that don't appear to come from a small source and grading, for the most part.

 

evanmoses: my post workflow:

 

1) paint job: fix errors and artifacts, paint in specific shapes to control contrast (in that last image in this thread, check the area underneath the main wall front left and compair that to the image without post to see what I mean) and faking some badly-done shaders/textures. I also reposition entire objects and walls when needed. the cross-hatched bars on the windows (top right) were painted. good old line tool

 

2) adding textures and detail objects: sometimes I didn't manage to get a decent reflection or specular highlight in 3D, so I can add it here, distorting the image to fit the shape it's being applied to. (see the map image front left in that last render). occasionally I add entirely new objects (duplicating lights for example, using the clone brush).

 

3) colour-correction - I paint specific shapes to fit the objects that need correcting and go to work on them. I usually duplicate the entire canvass (once flattened) and blur it a lot, then apply more colour-correction to the whole thing, but adjusting it's opacity and blending mode (screen, multiply, overlay and soft light work nicely for most scenes I do).

 

4) effects - volume lights (if unable to do them properly/efficiently in 3D), fog, lens effects, distortion etc. DOF if needed (sometimes done with a z-depth pass if I have one, or just hand-painted using the blur brush or using a blurry shape painted over the area required for blurring.

 

5) atmosphere - faked fill lights etc usually done with very low-opacity blurred shapes. grain (extreme contrast to get individual small clumps of noise) and finally grain/noise to match the background plate or just to make it look a little more atmospheric (perhaps to simulate a photograph more accurately).

 

6) go back to 1 and check everything is ok. repeat until you need more coffee.

 

 

hope that helps.

[Iain Denby]

Alex

 

Your post workflow is Ideal for making your work stand out. I do similar, and I too feel guilty sometimes for not being a 'purist' when it comes to 3D. Embrace it, don't feel guilty...there is too much work which looks the same around in my opinion. Animation is a totally different animal. deal with that in a different way if, and when it's required.

[Alex York]

Cheers Dibbers.

 

I've been using this workflow for about 3 years, now, and it's slowly becoming more and more refined. I don't feel guilty in the slightest, but I do feel a tad worried that when it comes to animation I'm gonna be completely up the creek! I have a freelance project with an animated archi scene in it coming up very soon, so that'll be the real test!

 

Cheers and I'm glad you like my style.

[Alex York]

Right! Last one...

 

 

This one was something slightly off-topic, mainly for practice, and is an attempt (stylised of course hehe) at recreating the Naoshima Contemporary Art Museum in Japan, by Tadao Ando.

 

Now for the 10,000 word report on the entire project...

 

I'll probably re-post this whole lot in the Finished work forum, when it's all been packaged up and finalised. I'll be retouching a lot of them, too, if I get time.

 

Thanks for the feedback everyone! It really helped a huge amount.

[evanmoses]

Hey Alex,

 

thanks a bunch for sharing your process. The last image is amazing. I'd say it's your best one so far, but that's not really objective as I'm a little obsessed with Ando's architecture. Still, great job, the floor is so perfect.

[Alex York]

Cheers Evan, glad you like it.

 

I discovered Ando's work only last week, so I feel a bit cheeky trying to emulate it already! Amazing work though, particularly his attention to light and how it plays off against large expanses of wall. The floor... I'm not totally happy with it. It could do with some blurring near the far end of the building, since the pattern ought to be less visible nearer the horizon. Should be easily fixable in post. I'm not printing any of these until I'm absolutely certain they're as good as I can get them (which is never, I suppose!).

 

My priority now is the report. Can you believe that the bloody weighting for this project is 70% report, 30% images? Unbelievable... That means I could still get an A grade even if I didn't do any actual renders at all! (although producing a report would be difficult without anything to discuss besides reference!).

 

Cheers again, and I'm glad you found my process interesting.

[Ernest Burden III]

I too feel guilty sometimes for not being a 'purist' when it comes to 3D. Embrace it, don't feel guilty...there is too much work which looks the same around in my opinion.

 

I went through all that with not being a 'purist' with watercolor, as I pushed the CG content of my 'watercolor' renderings to 50%, 75%, 85% 90%+ It's the final that matters, and not important how you got there. Make it however you see fit.

 

As I pointed out once (or twice) before--Photoshop is the 'money app' for most of us. Without it we are left with whatever the rendering app spit out, and it often isn't all that special.

[Ernest Burden III]

Ernest: the kind of post I mean (and I guess Tim means too) is paint work....

post for this sort of work is limited to colour-correction, large effects that don't appear to come from a small source and grading, for the most part.

 

 

I agree that you can't effectively re-paint portions in post on animation frames. But I do most of my animation post-processed to look very different from the original renders.

 

Two points:

 

First--identify what is needing the most post work, then go back and fix it in the model or rendering settings. Or live with it. Motion relieves you of many a visual problem that you would never leave alone in a still.

 

Second--take a typical image into Photoshop at DVD resolution and start recording an action. Do only things that will apply to other images like color corrections, curves, etc. Record everything you do. You may find that your process is such that it will work in a generic way. Once you have a recorded action you can batch-process an entire folder full of images (COPIES of images!).

 

 

I like the last image. But with the size I have my browser, I saw it differently, and prefer 'my' cropping. Also, the picture is about 1% off vertical. Why not stratighten it up?

[Alex York]

That actions method is really very useful. However, my process in post is usually dominated by experimentation, so producing an action would probably involve about 100 different steps, which would perhaps not be such a wise choice! I might give it a go on the next project, though.

 

Your cropping is interesting. I like it too, but it took bleeding ages to get those windows looking ok, so I'll be keeping them for now! I don't quite get why your image is cropped at all though... if you're working at a different resolution, it would only be squashed/stretched. Unless you meant that it was squashed/stretched, and that you also cropped it afterwards?

 

That vertical shift was intentional. I didn't want another standard tripod shot look to it - something a little more natural, as if it were being viewed from the POV of a person. Something I'm trying to introduce into my work lately.

[Ernest Burden III]

That actions method is really very useful. However, my process in post is usually dominated by experimentation, so producing an action would probably involve about 100 different steps, which would perhaps not be such a wise choice!

 

My actions encoded my experiments, so they still have some false steps that were un-done. I regularly use a few, and they ARE lots of steps. Maybe 100.

 

That vertical shift was intentional. I didn't want another standard tripod shot look to it - something a little more natural, as if it were being viewed from the POV of a person.

 

It's an old and often-discussed issue. But I don't believe that a person experiences the vertical perspective when they look up, especially if its only slightly. There is more to what we see than what we see. CG is more the latter. Your work in general is working towards the former.

[Alex York]

i'd love to know what you mean but "there is more to what we see than what we see" doesn't make much sense to me

[evanmoses]

Hi Ernest,

 

Not to criticize the master, but I think I prefer Alex's cropping to your version. With the notions of symmetry and the 'move' (physical and conceptual) from inside to outside being as important as they are in Ando's work, I think that including the right side gives you a better idea of the architecture. But I might just be letting ideas getting in the way of images.

 

Evan

[Alex York]

With the notions of symmetry and the 'move' (physical and conceptual) from inside to outside being as important as they are in Ando's work, I think that including the right side gives you a better idea of the architecture.

 

that's very true. I love how he balances natural light against man-made light (the exterior (blue tones) with the interior (yellow tones)). I tried to get that across in this render. I think I could have pushed it a little further, though.

 

Truth be told, I just got a little tired of portrait renders and went for a widescreen landscape this time

[RyanSpaulding]

So whats the difference between Final Gather and GI? Is it a similar system.

 

I must learn a new rendering tool. MicroStation simply does not cut it.

 

How easy is XSI to learn the rendering specifics? Does it handle importing DWG/DGN well?

[Alex York]

Well the FG/GI question is a good one. in all honesty I don't know. all i know is I can achieve the same look as high quality GI with FG and ambient occlusion, but in half the time and with far less setup. (GI in XSI requires a lot of setting up per object, FG just needs to be turned on and tweaked).

 

a lot of people use FG on top of GI, to enhance the effect and resolve flickering problems (since it's relatively fast, so you can up the samples and get a smoother, more detailed result) and then push it further using AO to get even better detail.

 

I've never tried importing DWG or DGN, so I can't help you there either, I'm afraid!

[Ernest Burden III]

I might just be letting ideas getting in the way of images.

 

What are images without ideas?

[Alex York]

good point

[IC]

Nice work Alex-it's coming along really well.

 

I think the reason for more people not using XSI for architectural work is the lack of DWG/DXF import. Bit ridiculous that something as sophisticated could be lacking something so essential.

There are ways around it but the results are frustrating enough when you can import directly.

[lambros]

XSI imports IGES and EPS files.

Saving EPS through Autocad though does not work directly, you have to load the exported EPS file in Illustrator and export it again as EPS Legacy (choose Illustrator 8 and uncheck the preview option). That works fine and is free (if you own Illustrator)

 

One other method is to convert DWG to IGES via a 3rd application (such as Rhino - but why do we have to buy Rhino to do that!@#!).

 

Another method is to buy Polytrans from Okino, which is a 3d model converter. There is a specific Polytrans for XSI plug in, that allows for direct import of many formats from within XSI. I think it is at $395 or something.

 

On XSI base, there is a wish list for XSI 4.5. Needless to say that DWG import is high on that list!

[Alex York]

Thanks Iain

 

I've been using FBX whenever I've had to exchange between apps, but yes DWG would be great! I'm not even sure ACAD or any of the major archi apps have FBX support yet anyway. openEXR is becoming pretty popular now, too.