I found this incredible render by Simon Thommes and had to post it here. The attention to the details, lighting and composition is outstanding. He even added a hair particle system to the cables!
Would you guys spot the CG part of it if you randomly found it on the internet? If you would, let us know what gave it up in the comment section below!
9 Comments
Depth-of-field grain blur, but that's a byproduct of the rendering engine and not a fault of the author.
Kind of looks like they manually rendered the DoF into the scene itself rather than adding it in post, which is a shame if they did... Otherwise it is a cool render!
I think, the grain is kind of an artistic effect here, cause even if it's not, it could be easily removed after the render is done. The render engine's DOF is a must have if you want a correct one. I bet you cannot get correct DOF on post anyway, except using deep compositing, which Cycles cannot output.
Yes, we all know how to put depth map into Defocus node, but it gives incorrect results on the edges of objects. For that we need a depth information for each pixel, what drastically enlarges output files. For more information google Deep Image / Deep compositing.
I'd love to have some hack for Cycles to be able to approximate defocused areas somehow. For now it's nearly impossible to get superclean DOF.
You know that was a ridiculously pretentious way of getting your point across.
The only counterpoint I really feel the need to say is that while a raw depth map doesn't create ideal results by itself, it allows dynamic controls which is quite important for some as it saves a lot of time despite a larger file size. The depth map can be combined with edge detect nodes and similar effects to refine the edges a solid amount.
Not to mention file size is a null pount for still images in this day and age. But I'm just basing this of what I've been told by my lecturers on this matter.
That being said, better pre-render defocusing could be a boon for some, depending on their style or approach. And yeah, the grain could be for artistic effect. Gven the style of the model, it seems quite likely.
Pretentious? My point was about not blaming the author for the real DOF. Dynamic controls over DOF on post is quite obvious, but once again, you cannot get correct results pixelwise with it despite all the tries with depth map and edge detection. No offense to your lecturers, but the only way to get near-correct results is to manually divide image into depth layers with masks, but you'll get step depth graduation. Or as i said before using deep image for depth map.
There's no reply button on your other response, so I'm putting my reply here.
Yes, you original comment comes off as pretentious. Especially the second paragraph. I'm guessing it's not your intention, but the wording really gives off that vibe.
I'm not sure how DoF works in blender, as my specialty is 3ds Max and NukeX for comp, so our experiences may be different, however, using depth passes is an industry-standard technique. For video, at least. Given that my Lecturer has spent quite a few years in the movie industry, I tend to trust him on this.
While I could have worded my original comment better to establish it's based on my personal views, I had no intention of "Blaming" the author for the effect.
Deep image compositing does sound pretty interesting, however, the availability of renderers able to actually output multiple Z-passes like that seem few and far between, not to mention expensive. The only easily available option seems to be Renderman.
This is trailing off, but I get your point, but with limited resources, I would much rather use a z-pass and spend time trying to make any artefacts as unnoticeable as possible, rather than having to re-render an image/sequence if the DoF is not correct.
You have really intrigued me in deep compositing though, sounds like something to look into over the summer!
You are completely right in this comment. I might sound a bit pretentious because this is not my mother tongue, so sorry about that. And i'm talking about the topic from my experience in this field of working in CG movie industry for four years. We're are using Renderman, Maya and Nuke. And as Pixar released path tracing engine we started to use real DOF. Never used deep compositing by the way, so all the troubles with DOF went to our compositors before. Renderman has a denoiser. Hope Cycles gets a decent one also in the next release.
Offtopic: Blender + Cycles shading workflow is the best realtime shading/rendering systems i've seen. And it's a pure pleasure. Renderman is waay off in terms of flexibility and usability.
His original reply was not pretentious, but rather anticipatory with future comments challenging his observations.
The fact he's correct makes your attempt, Ashley, as nothing more than deflection: a useless waste of time.
Now that is pretentious, but with conscious intention, on my part.
Technically it is fine (needs more samples - this gives it off), it is definitely in the area of realistic renders. Images with strong filters are always hard to tell what the source is.
But artistically it does not speak to me. I give hard super-honest critique nothing sugar-coated:
- there is not enough interesting things to focus in the piece - it feels a bit boring. In first image there is only a corner and cables in focus. Maybe the color palet is boring, maybe because it is almost monochrome because of the overuse of filter(s).
- the last piece has some composition though out, but the first two feel like random shots.
- there is no story, no interesting lighting, nothing to guide eyes over the image. The mic blends with the background like it is hiding from me.
I hope this was constructive for the author, if it didn't come as such, I am sorry for being a douche:)