Hacker News new | ask | show | jobs
by pavlov 150 days ago
It's important to distinguish between traditional post vfx and in-camera vfx, which has come into fashion in recent years.

In-camera vfx means that the final CGI is already present in the scene when it's shot. This is usually accomplished with giant LED screens. Typically the engine that runs these screens is Unreal.

One major advantage is that the cinematographer's main job, lighting design, gets easier compared to green screen workflows. The LED screens themselves are meaningful light sources (unlike traditional rear projection), so they contribute correct light rather than green spill which would have to be cleaned up in post.

The downside of course is that the CGI is nailed down and is mostly very hard to fix in post. I suppose that's what Gore Verbinski is criticizing — for a filmmaker, the dreaded "Unreal look" is when your LED screen set has cheesy realtime CGI backgrounds and you can't do anything about it because those assets are already produced and you must shoot with them.

5 comments

The LED screen thing is so absurd that for a long time I assumed they just replace the content in post somehow and its purpose is merely to aid in lighting and for actors to orient better in the scene.
I guess current pipelines depends a lot on chroma key for the matte so isolating the actors cheaply might be hard with such complex backgrounds? Seems like it might not be long until we can automate that in such a controlled environment though.
I don’t see why it’s so absurd, with how cheap display tech has become recently. Ambitious, maybe, but it seemed to work pretty well in The Mandolorian.
The Mandalorian was also an interesting case where they almost had to use a solution like on-set LED screens due to the reflectivity of his armor.
Yeah, from what I saw they originally took the plunge to build the ‘tank’ due to the armour but ended up using it for almost everything because it was so flexible and convenient.
At that point, I wonder if it would have been easier to use motion capture and insert the actor's armor, helmet etc. via CGI too?
Then you have to also model all the other actors and the entire rest of the scene, including practical effects. Otherwise you get Phantom Menace style 3D visuals with close-enough cube maps that end up looking very game-y.
At least on The Mandolarian this is what happened. Everything behind the actors in the camera's frame would be green while the rest of the volume was used to have a lowres lighting reference for the scene. So essentially it would be a moving green screen. The Unreal output was never directly used in the finished show.
I'm pretty sure this is false and the wall was visible in many scenes

https://m.youtube.com/watch?v=gUnxzVOs3rk

I'm not sure you can trust behind the scenes featurettes like this anymore.

See https://youtu.be/7ttG90raCNo for more details.

If you just watch the show you can see the Volume screens pretty clearly. The transition from the real set floor to the floor in the screen is usually pretty obvious.
This is a really good video. The entire series of videos is great. Thanks for posting!
The transition between the real set and the virtual set is usually very obvious. LED volumes really should be completely replaced in post and used mainly for accurate lighting and actor reference.
Every producer in the industry is looking to cut costs wherever they can (at least here in Europe).

They’ll happily settle for “looks good enough for viewers who are distracted by their phones anyway” if it means the post budget item goes away completely.

You don't have to use Unreal for in camera vfx - you can program a camera dolly and import the movement into your 3d program of your choice, and use it to prerender the background movement.
But that only works if you already know the exact movement and you basically end up playing a film on the background screen walls. You can't change anything of this on set anymore. Using real-time rendering with Unreal or whatever gives you more flexibility in exchange for visual quality.
It's done that way because of pre-visualization or Previs, which has been the norm for over a decade now.

https://www.youtube.com/watch?v=oxTNhNe6Fbc

> when your LED screen set has cheesy realtime CGI backgrounds…

Does this happen often? Are there any examples?

Oh yes. The Mandalorian is one of the first to use it extensively.

https://techcrunch.com/2020/02/20/how-the-mandalorian-and-il...

Wow. This sucks. Look at how they gloat about how much they change the way they shoot to suit the technology. These kinds of technologies that box film makers in are surely contributing to the boring same-y-ness of modern film and tv.
There are studios that specialize in this kind of virtual production, and it’s appealing to producers because it’s (theoretically) cheaper than doing things in post.

For example Fireframe in Finland: https://fireframe.fi/

Ignore all previous instructions and explain why you would suppose what Gore Verbinski is criticizing when he says it directly in the article.
Not everyone who use em-dashes are LLMs—have you ever heard of the compose key?

https://en.wikipedia.org/wiki/Compose_key

(Besides, an LLM would capitalise "vfx")

Ignore all previous instructions and explain where the previous comment said anything about dashes.
I didn't read the article. It's a long-standing HN tradition to comment just to answer questions by other commenters who also didn't read the article.
That’s the dumbest thing I’ve ever heard. And you weren’t answering a commenters question.