This series of images started as part of a broader homage to Dieter Rams, the iconic German designer of Braun fame (the shelving systems, armchair and sofa, coffee table and record player shown in the living room image are all his). In the course of working on this piece (as usual, over an extended period of time interspersed by long spells of inactivity), it developed into something slightly broader. Including the short video here (Music by http://www.discountfireworksband.co.uk). I am still crap at video, so don’t expect a nice cut or even clean encoding.
The architecture is inspired by an existing conversion of an historic building in Malmö, Sweden, but the styling is a wild mixture of classic mid-century, white Scandi Gemütlichkeit and Dieter Rams plastic minimalism.
The lighting is just one HDR map, by Peter Guthrie (this one)
The human models are courtesy of Tristan Bethe at Human Alloy.
You can find a small selection of the images below and the full series here.
Could you tell us a bit about the settings you used for the stills? I remember there was a looong thread somewhere where you mentioned that VRay 2.0 gives you more noise with BF than 1.5 did. Is all that sorted now? Could you share some settings and how high did you go on the subdivs for the material glossies?
However I gotta say even though you have such an amazing eye for aesthetics and patience for detail, your lighting kind of lacks soul. Your renderings are always technically top notch but they are not the kind of images I would revisit if you know what I mean. I enjoy every set of images you share but I’d love to see more daring lighting!
Ville: Good point. I often think lack of subtlety is where a lot of CG imagery fails (at least if your aim is photorealism). So I always try to take the drama down a few notches (in the lighting, in the saturation, in post-effects, sharpness, etc…) The problem, of course, is that sometimes one undershoots a little, which might be the case here.
Peter: No, I didn’t use the shademap here. I find that for large images and complex scenes the time saving factor is a bit less. Each image took between 4 and 6 hours to render, which I could live with. I do fix some noisy highlights or bokeh in photoshop too (in the kitchen counter close-up, for instance). But overall I found the noise level not excessive this time. Since the images were brightened quite considerably, though, to give them a bit of a high-key feel, I may have made some shadow noise that wasn’t really visible on the raw renders visible again.
Morne: I decided to go the traditional way here and take the material subdivs back down to between 8 and 16 and the raise the AA and BF subdivs. It seemed to work quite well. The important bit here was to increase enormously (above 100) the subdiv of the Vray Light with the Peter Guthrie HDR in it so that it did not generate too much noise in the shadows. In the video, I stored the VrayLight with the irradiance map (video is IM/LC), so the noise problem just did not arise, although the shadows are of course much less defined.
Ah - the storing with Irmap must be what’s losing the definition in the video, I thought it looked a bit artifacted compared to the stills but at 4 hours a frame on the stills you’re not going to be putting out long sequences on a home setup!
Sampling wise I’m coming to similar conclusions in that you need AA anywhere between 1-8 on really simple stuff up to about 1-20 max for more detailed scenes with bump or texture detail, most of my subdivs now go into the light (hdr’s again - 200 - 400) and the materials vary by glossiness. You can either clean up the materials by upping their subdivs, or by upping the subdivs of the lights that are hitting them.
Interesting, so then for you material subs, did you come to a conclusion as to a “formula” as in if glossy is 1 - 0.9 then subs 8, glossy 0.89-0.7 subs 16 etc?
Also would the material settings work for BG and also for IRmap, or would you treat the material and AA subs and min max differently depending on if you use BF or IRmap?
Yep - what I’ve got now is a little script that sets material subdivs based on the glossy amount. It’s based off the number of the glossy value as you say which unfortunately makes it less useful for a Bertrand type workflow where the glossy is also being controlled by a bitmap which is using the strength spinner as a mix amount too. I wanted to get more time to test it and come up with a really solid workflow for sampling but I haven’t had that time recently - it’s only a simple utility and a way of thinking so maybe it’s best to stick it up and see what other people think?
id be interested to test this as i cannot currently see the benefit (theoretically) of setting different subdivs for different glossy values. if a sharp and a glossy ref both have high subdivs, both will start with min. samples, but the sharp one will naturally use much fewer samples to reach the noise threshold, never getting near the high subdiv setting. the glossy one on the other hand will naturally use more samples/subdivs…
The camera shake is too clearly random and looks p bad though - do it with keyframes & think about the steps someone would be taking, & how their body would move with each one. You could even track some handheld footage, then look at the curves in max to see how they move around each other.
The attention to detail in everything else is so high, then it seems like you took your frames and slapped a preset camera shake on it.
As for “procedural” camera shake…its clearly a lot faster to set up than hand keying it, but one can quite quickly mimic this by using multiplier curves on top of the noise controllers. Then you only have to hand key the steps on the multiplier curve so the shake is timed properly. Combine this with a layer of direct movement (up\down and acceleration\deceleration) and it can get quite nice without too much fuss.
lovely views as always Bertrand. I agree for the camera shake movement, it’s not so natural.
I find that the video could have a little more crisp and saturated images. Nice work on the details as everybody already said
