Century Plaza

[squintnic]

ah ok interesting, we have AE artists but not thought of doing it that way. many ways to skin a cat!
we haven't had many issue at all with F9, using a mix of i9 and Xeon with 2080 series cards mostly. Neat video in OFX is pretty amazing (and fast with GPUs) for denoising / deflickering in fusion for us. What are you using to tame hot spots and over bright pixels?
Where are you doing your lens fx / glows / flares in? They always look nice and considered.

[fixeighted]

I actually used AE extensively to debrand hundreds of F1 video shots and whilst it was a pain I developed a workflow and techniques to make it all pretty simple.
AE definitely has a place but it's all about what an individual is comfortable with I guess. At one time of course it was pretty much the only option for basically everything....now there are options....all good
I think that projection technique may have helped in some case but a lot of it was just brute force, frame by frame stuff.
Mocha I've used a great deal over the years too; very cool and ideal for a lot of specific stuff. It all depends on the particular shot what feels best to use.
I would probably have started with Mocha for shots like these, area planar tracks with custom patches based on clean plates...stuff like that.
Although thinking about it, if you've already got a solid track then much can replaced with geometry and comped in.

[Neilg]

ah ok interesting, we have AE artists but not thought of doing it that way. many ways to skin a cat!
we haven't had many issue at all with F9, using a mix of i9 and Xeon with 2080 series cards mostly. Neat video in OFX is pretty amazing (and fast with GPUs) for denoising / deflickering in fusion for us. What are you using to tame hot spots and over bright pixels?
Where are you doing your lens fx / glows / flares in? They always look nice and considered.

Bright pixels was a mix of techniques - sometimes just isolating the overbright pixels as a mask and nudging the clip over by 1px underneath it... dirty, but it works when there are a lot jumping around.
glows and flares are a mix of techniques - we exported a vray glare pass, in some cases hand placed and tracked them, have the exponential glow fusion macro and used video copilot - all very subtly layered on top of each other. the flares were applied in AE in layers just before the grading / contrast, even though some were generated in the fusion file.
For hot spots we render with highlight burn set to .3 but dont bake it in - so that in fusion we use a mixture of highlight compression and reinhard and can mask them as needed to get the look we're after. a little fiddly but you have so much control, it's worth it. also important for glows to be generated from the pre-compressed raw render.

I cant believe 9 is stable for you! any machine at DBOX, you can open it, leave it alone for 10-15 minutes and it will crash.

[squintnic]

thanks for the explanation - im going to use that 'dirty' method right now for fireflies.
we also render linear for glows and then compress highlights in post. Flares for us are with Sapphire but it can be so touchy with 32bit exrs and blow out immediately, meaning annoying trips deep into the menus to tame the blow outs..
really weird you have such consistent errors with fusion 9

[elivnA]

Very well done! Ofcourse now follows the most important question: was the inflatable flamingo a bought asset or made inhouse?

[100%indifferent]

Great stuff as always Neil from you & the Dbox team. I'm wondering how you guys handle rendering; I'm assuming you're not going the gpu route & are still cpu/bucket. Do you render in house, cloud, etc? Lastly what type of settings; brute, imap/lc, etc? I'm also assuming you guys are on Next, I'm still on 3.6 & a bit reticent to upgrade until the online licensing hassles are sorted. Any details you can share about your flow are appreciated. Thanks.

[Neilg]

Great stuff as always Neil from you & the Dbox team. I'm wondering how you guys handle rendering; I'm assuming you're not going the gpu route & are still cpu/bucket. Do you render in house, cloud, etc? Lastly what type of settings; brute, imap/lc, etc? I'm also assuming you guys are on Next, I'm still on 3.6 & a bit reticent to upgrade until the online licensing hassles are sorted. Any details you can share about your flow are appreciated. Thanks.

Thank you
for this we were all brute force/LC. did not have time to pre calculate anything. render times were an hour a frame target.
we operated on a very strict render process - all client review and approvals were done using first and last frames only. once those were approved alongside a max viewport camera path we tested 5 frames to make sure we didnt have flicker and then dove into final render. 70% was done locally on 6 machines - 3 ryzen 2700x's and 3 i7's that took twice as long to render. almost every sequence you see in the final film was the first time it was rendered. 5 of the machines were workstations so all rendering happened at night and weekends.
We had final sequences rendering 4 weeks into the 10 week project, it was military level render scheduling.

the other 30% of it was rendered on amazon ec2 - the opening exterior and lobby shots, but it was miserable. it took a while to get those approved and they needed to be turned around too quick. they were also the slowest and would've been 3 hours a frame locally. most of the time managing that was out of office hours and weekends. it was horrible. we billed the client for what it cost but if they'd approved a week earlier and we had 2 extra boxes, it would not have been necessary.

edit: yeah, no gpu. we use studio scenes set up with gpu for asset/material development, but all major complex scenes are cpu. cant risk failure.

elivnA : thanks! we bought it from 3dsky, but it was a last minute addition, those scenes came together 2 weeks before the end. we'd gone through the whole storyboard spending a lot of time at the hotel pool, yet once it was rendered, visually it was just boring to look at. we are very happy the client was up for it because if they fought back we didn't have a plan b!


EDIT: just realized I have this uploaded. here's the first draft storyboard we got approved by the client!
https://www.youtube.com/watch?v=q9Wg...ature=youtu.be

[100%indifferent]

Neil, thanks for the info, so interesting. In my mind, after having followed Dbox's work for years & having read your posts for equally as long, I'm shocked at the hardware you're producing these fine films with. I had imagined Dbox had a big server room chock full of blades just churning frames out. Just goes to show, you can always get by with less. I'm a one man arch-viz show, though my farm is a 5 node Boxx dual-xeon setup, and when I'm rendering some animations I always feel like I need/want more speed/power. I'm in the market to buy a new node & have been looking into the gpu route as it seems its inevitable that it will replace the cpu route at some point. However, the stability & predictibility of the cpu route hasn't failed me yet, and like you guys, I don't have the luxury of risking failure. Some of the issues I've read through on the Chaos GPU forum still make me feel like we're not there yet for heavy interior arch-viz animations just yet.

Anyway, thanks again for the intel, it's always appreciated.

Looking forward to your next share

[AlexP]

Looks great,

The intro though.............

If not too nosy, I'm really interested in the numbers, how many artists, hours, etc. how many artists doing 3D only, how many doing other stuff? (post etc)
Did the client happily sign off stuff as you went, or did you end up with a big crunch at the end after the client had made the obligatory last minute changes (of mind) ? (because they didn't notice it in the previous 100 drafts)

edit; just noticed 5 people. 10 weeks, still interested in whether that was 5 people doing 3D or if responsibilities were split?

[Neilg]

The intro though.............

If not too nosy, I'm really interested in the numbers, how many artists, hours, etc. how many artists doing 3D only, how many doing other stuff? (post etc)
Did the client happily sign off stuff as you went, or did you end up with a big crunch at the end after the client had made the obligatory last minute changes (of mind) ? (because they didn't notice it in the previous 100 drafts)

edit; just noticed 5 people. 10 weeks, still interested in whether that was 5 people doing 3D or if responsibilities were split?

Yeah I nearly didnt share it because of the intro...
cant win em all.

the client signed things off as we went - we were fortunate that the client was passionate about the film and cut out a lot of people so his word was often the only word we heard. we didnt have a sales team and brokers and the interior designer and architect all having 'opinions' about the representation. helped a lot by the project being pretty far along and under construction already.
There were a few little scambles when information wasn't arriving and we had to threaten to miss delivery and charge more - then we got what we needed within 24h. It always existed, it was just difficult to find.
We had a timeline from the start which got refined weekly as we worked, really helped keep the train on the tracks.
There was a twitchy moment where the client told us that the lobby was wrong, so we got new information and then he said we'd gone even further away from what the lobby was supposed to be. turned out the design we got was what was being built - the stone had already been ordered and everything. someone had disagreed with him and told the interior design team to just do something anyway. that was a wild moment... he was fucking pissed, but ultimately we did what was in the drawings.

all 5 people did 3d, myself an another guy handled the direction & edit / postwork experiements / cloud rendering / client calls. Matthew (ceo of DBOX) was also heavily involved in the direction side of things and client management/pitching each stage. we did 110 hours of overtime total (thats everyones added up) during the whole production on top of 9 hour days full time. That was pretty evenly spread - the last 4 weeks I was staying until midnight twice a week managing renders. Nothing at the weekend beyond remoting in and making sure renders were going ok.
Funnily enough thats about the same amount of hours that people had booked off as PTO for that period of time - I had a full week off in the middle. We managed to keep the process fairly stress free - there was a mountain of work to do, but we didnt have any major diversions that couldnt be solved with a few extra hours.

[AlexP]

Thanks for the insight. So important being able to keep client in line with your own timing.
Always great getting in the middle of some design politics hah

[AJ Jefferies]

Holy shit man, this is really impressive! Amazing work.

[Neilg]

Holy shit man, this is really impressive! Amazing work.

Thank you! means a lot

[PIXELBOX_SRO]

Amazing as always Neil.
wondering how the heck were you able to roto out al those cranes you described... that must have been mental and as much as i have watched it i could not see any leftovers there

[AJ Jefferies]

"3 ryzen 2700x's and 3 i7's that took twice as long to render"

I just spotted this and had a question - Do you have any issues mixing AMD and intel CPUs? I've just upgraded to an i9 and was going to opt for a Threadripper but after getting burnt by the floating point issues on AMD many years back I was still too wary to take the leap.