A Western VFX guy in China

Author Archive

VFX Talk Thread about the Chinese VFX Scene

Hi there,
 
i just want to recap forum activities on VFX Talk especially my thread about the Chinese VFX scene (Window into the Chinese VFX Scene) which caused a lot of turmoil but also attracted many interested readers since there was no real information available before this entry, at least it seems so. Especially in times where outsourcing is a ‘buzz’ word and many fear that their jobs in the West will be replaced with the ones in low labor countries like India or China.
 
If you have time please feel free to read through my really long blog entries to shed some light on the domestic scene here in China and also understand the obstacles to overcome if someone is willing to do their first steps in China. Until now i have seen many western setups failing and unless founders and operational executives have gone through rigorous management training in how to deal with Asian counterparts and how to cope with randomly and ever changing regulatory frameworks, failure is most likely.
 
Advertisements

Xi You Ji – Monkey King TV Series

Late entry, but i was too busy from Sept. until now to blog anything. At this moment i involved as on-set supervisor for the new Monkey King TV Series (50 Episodes) which is expected to go on-air with the beginning of 2011. With an overall budget of nearly 100m RMB (14.5m USD) including post-production the highest budget ever paid for a Chinese TV series. Equivalent shows were made for a 10th of the costs in previous days. But the producer Song Ya Ping, shareholder of the Huayi Brothers and highly acclaimed TV producer in China aims at a LOTR version of Monkey King and should prepare him to pitch a 100m USD version of the same saga which is one of the 4 most famous novels in Chinese literature. 
 
My job also includes the customization of the 4 F900R’s meaning the creation of a film look with LUT’s and advising on the whole post-production workflow and DI. I also built a on-set previz system which is able to do high-precision keying in floating point space and can simultaneously record up to 3 uncompressed HD live feeds.
 
At this point in my Chinese career i just want mention that improvisation in China is everything and that things usually get done much faster than in the West despite the accidents that occur from time to time like a broken crane where a camera operator dropped 15 feets (he survived 😉 ) or a camera get smashed from the wire crew or explosives were triggered to early, etc…
 
Eventhough western crews would probably not cope with the non-existing safety standards or working hours, producers are normally very pleased with the results from Chinese crews which are tough like bricks and no matter weather or conditons hold their positions until the last scene is finished. Much respects to those guys, especially to my Hunan Kung-Fu instructors and their stunt crews which are undoubtfully the best i ever met.
 
These photos are just few impressions about some locations, we shoot more than 40.000 pictures until now and later entries will introduce stunning vistas shot with
a 16 Nikon camera rig from remote locations like the New territories, Xin Jiang, Xanxi, Tibet, Yunnan and others which will be sold later to Getty Images.
 
 
  

Bubble Gum Crisis Wrap Up

Ok, this entry should only act as a wrap up of our finished Bubble Gum Crisis trailer we did for Cannes. Unfortunately not what we expected but still interesting to test out my RED workflow with FCP and Color (aehmmm..) Anyway for short-term projects importing in FCP in RED’s native colorspace (Log&Transfer) without transcoding to ProRes 422 (HQ) might be a performance hit, but for our 2min. long timeline in 2048×1024 worked pretty much flawless on a 8core machine. No frame dropping if playback quality is set to Medium, good enough for editing and still be able to see all necessary details of the acting performance. 
 
Regarding our film workflow, i sent to Color and made a one-light correction as well as built a custom LUT to save a DPX sequence in LOG space. Was not so difficult took me some days and with the help of my recorded Mac Beth and calibrated color charts wasn’t really rocket science. Why DPX ? is 12bit color-precision not good enough of RED code, No? just because we are all used to DPX workflows and also to do a quick print out on a Arri Laser recorder which expects LOG space anyway.
 
 
 
In Nuke i just applied a standard LogLin operator switched to ‘Linear’ viewing LUT and everything was just fine. The CG were rendered in Maya with Renderman and we saved everything in the OpenEXR format and recombined the passes with a custom written Gizmo. Isn’t that complicated at all and we frequently do this with VRay passes from 3DSMax. Maximum flexibility and artistic freedom for the compositor. My prefered workflow consists of the OpenEXR format in Nuke rather than getting 30 different layers as separated passes, sometimes as TIFF’s which is not only a waste of storage space, but also a really hit for CPU performance. Is just for print, for god’s sake not for comping, OK ??
 
The comp was pretty much straight forward, just the keying needed a bit of love, but with a combination of the new IBK Keyer and of course my all time favorite the Primatte and Keylight (hard and soft method). Some roto to remove rigs and some paint fixes for the wire removal, but nothing dramatic. After staging (framing) animating the Maya camera, exported the cam data with a Mel script to Nuke and built a Pan & Tile for the skydome, tweaked the 3DMoblur with the provided velocity data out of the scanline renderer, smudged up and de-saturated everything, put grain on and rendered out. The really cool thing though is that the TOD (Time of Day) could be preserved throughout the compositing stage which meant that i was able to do a quick confirm with the EDL from FCP in Color. Since i didn’t have Gluetools  to import the DPX sequence in FCP, i rendered with an uncompressed 10bit RGB Kona3 codec to preserve the latitude. Anyway FCP as onliner not really an option, but with some workarounds doable. Would have been much easier with a Smoke but there was no budget for this.
 
The lighting was sometimes off, so i had to play a bit with our RGB Diff pass, but for a trailer and in a hurry i guess, OK. Not the best trailer i did in my life but again a very important lesson in understanding the Chinese way of doing things (aehmmm…at least i tried hard
 

Bubble Gum Crisis Trailer

At this moment we are dead busy with a trailer for Cannes. Bubble Gum Crisis is a successful Manga TV series which has been airing worldwide for the last 20 years. Now the Japanese copyright owners decided to make a feature film adaptation and try to sell the whole package at the Cannes Film festival this year. Since time is short, resources limited we had to very careful explore all the options available in letting this to become a reality. We have been brought in at the conceptual stage where the actual designs and the transition from 2D to 3D and further to real characters were not even decided. The production and the conceptual stage are running in parallel which means a logistical nightmare for the pipeline as models, animations, textures have to be constantly tweaked to address the wishes of our client and to ensure that the fan community will get what they would expect from a comic (Manga) adaptation. Our pipeline consists of Mental Ray for preview renders, Renderman Studio for final renders, Nuke for compositing, Maya for modeling and animation, Z-Brush for texturing, 3DSMax (Fume FX) for Particles and Houdini for FX animation. We’ve adopted the Open EXR standard for rendering and compositing and are pretty happy not to run into significant problemsuntil now.

Below you can see some production stills from the shooting with RED One and our ‘garage’ studio setup. The overall timeframe is 3 month from conceptual design to delivery. For a 1,5 min long trailer a pretty tight schedule… But hey, it wouldn’t be us if it would be easy !!

 

 


RED ONE SHOOTING

This one is special for me… i shot the first time with RED and even there are voices against that it can’t hold up to Genesis, F23 or others, i must say i am seriously impressed with the 4K resolution (i know just 3,2) and its nearly grain-free appearance. We shot 4K 2:1 to down sample to 2K 2:1 and after a simple one-light treatment in Cine Red (which is BTW impressive too) exported as DPX sequence to further do the final comp in Shake. I was a little bit puzzled first to be able to change basically everything from sensitivity (ISO/ASA) to White Balance, Tint, Temperature, etc. but once you get the concept of RAW it is pretty simple to get the a ‘look’ out of the box. Indeed a really geeky thing what RED throw on the market and for sure not everyone’s thing to take over the ‘colorist’s’ job as it necessitates more than just pushing buttons and a deeper understanding of the whole post-work-flow down the road, but for people who love ‘total control’ a heavenly dream came true.

For me for sure not the last time and we are hopefully soon going to shoot everything from TV-commercials to TV-movies and maybe act as second unit for feature films. Especially for cost-effective VFX shots RED is from now on my choice. Anyway, i want buy Scarlet when available and i am a huge fan of RED ONE, even if there are still some people concerned about reliability and matureness of such a system. For geeks like me who want be involved in the evolution of a revolutionary new digital cinema format definitely the right (RED) thing !!!


Nuke 5.2

OK, so now that i have got Nuke 5.2 and just finished my first comp, where i was using the 3D environment to displace a camera-mapped image to simulate footsteps of a guano, i must say that i really like Nuke’s channel architecture even though i think i am still much faster in Shake to these comps with a lot of separately rendered passes. Maybe because of the clarity of piping just in the mask layer for certain operations, which can be done of course in Nuke as well, but in the moment we rendered OpenEXR’s files it was sometimes confusing to choose the right control layer, you always have to switch back and forth to see what the actual layer was, whereas in Shake it was much faster to see which layer is used to restrict a certain operation. I think it is just a workflow thing and i have to get used to, BUT the rest i pretty much liked especially the 3D environment. 

We will also soon open a NUKE course here at IDD (www.iddcg.cn) which i will teach besides my successful Shake course and also help Base FX (www.base-fx.com) to make a smooth transition from Shake to NUKE.


The BEST will win…

So here we go, what should i say: a cool loft in Chaoyang district, i am their VFX Supe and Lead Comper with another 120 !! folks, doing tests for Rainmaker, The Orphanage and others, pretty impressive what kind of transformation company XXX went through in the last year, nobody probably expected to see that kind of work from China, at least not from Chinese guys, BUT if their US clients are already afraid of loosing their core business to China than they must be on the right track to gain international recognition and i am somehow one of the reasons to give them a more competitive edge.

Here are screenshots from my tests i have done for Rainmaker:                                                                                                                              

 

                   1. original live plate                                                            2. jaw removed                                                     3. CG jaw (face) replacement

I did the jaw removal and comp in Shake, facial tracking with rough gemometry in PFTrack and the animation and fur was done by company XXX (Shave and a Haircut) in Maya.