Epic's Distant Future

[Mark B.]

I suppose this thread is a question for Jim. What's the long-term plan for Epic? Years from now, when technology allows cheap SLR's to shoot 1000fps at full chip resolution, are the Epic brains going to be priced near those cheap SLR's? Or will you be selling 500,000fps Epic brains at $30k? I'm wondering, because if we are going to invest in replacement brains in the future, I'd like to think that eventually they'll become more affordable and not just live in the land of better performance.

[Martin Weiss]

The specs will always be subject to change, but you can be sure that RED will always produce tank-quality cameras that give the best images possible at that price range.

[David Rasberry]

One of the biggest differences between cheap SLR's and their more expensive brethren is robust construction that can survive extreme environments and rough treatment.
Lets see a cheap SLR that will survive a 6 foot fall off of a tripod and still work.

[Jaime Vallés]

One word: waterproof?

[Jerrod Cordell]

One word: waterproof?

That'd be pretty shweet.

Plus 20+ DR, 5000 ISO, the list goes on and on.

There's always room for improvement in technology.

[Jannard]

If we don't offer a price/performance advantage in the distant future, we won't stay in business. Our job is to lead the performance development and never get behind.

Jim

[Stefan Christou]

I'm not sure Red would have much problem with becoming the next Canon (and then some)

Edit: Jim spoke before I refreshed the page (by posting). But my point exactly!

[sergio arguello]

One word: waterproof?

done already!!!

[Jochen Schmidt-Hambrock]

Regarding noise floor or dynamic range of sensors the natural laws and physics are the limit. The faster the development cycles, the sooner we will reach hard boundaries.

You cant get better than 1 photon -> 1 photosite = valid signal, not noise.

Jochen

[Blair S. Paulsen]

It could be argued that the extreme level of price/performance advantage of the RedOne was so great that the Sonys and Arris of the world resorted to FUD rather than try and compete apples to apples. I suspect it would have taken them 2 years to get a "me too" product to market and that it would have been impossible for them to make any money at a sub $20K price point.

The announced Arri D-series cams are an obvious effort to tread the path RED has blazed. It would seem that they believe the Arri name justifies a higher cost and based on the typically bulletproof nature of their gear a good portion of the market may agree - particularly rental houses. What remains to be seen is how close the EPIC is to the Arris in build quality. I expect that Jim and the Red Team will have learned a great deal from round one and that issues like the battery mounts on the RedOnes will NOT be part of the EPIC experience.

Cheers - #19

[Stefan Christou]

1 photon would be pretty bloody sensitive!

[Eki Halkka]

There are plenty of places to go beyond sensitivity, frame rates and resolution...


...my current favorite, which should be doable in nearer term (even enabled on current cameras), is saving multiple exposures per frame - that way one can enhance dynamic range, and/or reduce noise or i.e. adjust shutter speed in post.

Plenoptic cameras are also a cool idea: they allow one to dial in focus and the amount of DOF in post, among other things.

Then there's HDRI, 3D cameras (not the stereo 3D variety, but actual 3D mesh creation / gathering depth information on the fly) etc. And then some, things that haven't been thought about yet.

...as said, plenty of places to go.

[Roberto Lequeux]

Regarding noise floor or dynamic range of sensors the natural laws and physics are the limit. The faster the development cycles, the sooner we will reach hard boundaries.

You cant get better than 1 photon -> 1 photosite = valid signal, not noise.

Jochen

"Color sensing sonar" type stuff. :)

[Wesley Scoggins]

I think the future is "Cognitive Cameras". With a powerful enough system, (like the human brain), we can take a signal that has major deficits (our eyes provide us with an upside down signal, full of holes, and crisscrossed with visible vessel structures.) massive amount of the visual data we take in is effectively "created" with our brain constantly filling in the gaps and limitations of our sensory structure.

So I think with a "smart" enough processing system, you can make major improvements on how the data is internally processed.

In a few decades when you have processing structure in your camera that are approaching the complexity of the human brain, they'll be able to do things with the signal that we can't imagine our computer doing now.

Like imagine all of the noise reduction and image sprucing that the guys at IMAX or Lowry are able to do for old damaged films. Imagine if you had all of those tools that they use to squeeze detail out of an image, and the quality of all artists they have working frame-by-frame, but you had that on a chip (and with a vastly superior source image to begin with.)

I think the future is not just Plenoptic camera systems, but taking that data so you can not only control the DoF after the fact, BUT taking the data and interpreting it as 3D data. I think that the data you collect while filming won't so much be treated as single frames, but systems in the future will be able to interpret them as images in a full space, and it'll be able to not only stitch together the space, but if you want go beyond the limitations of what you filmed, and fill in the "gaps".

So lets say you filmed a person running down a street, it could not only take the image of the actor and take it out and treat it as a 3D object, but it'll make educated guesses about what the other side of his face looks like based on what it sees of him. THEN if you want to reposition the camera you'll be able to do that in post, treating what you recorded as a 3D space, and being able to move your camera in it.

If you want you'll also be able to tag your video with GPS data, so it can download High-Res images from Google Maps, and if you wanted to pan around and film the other side of the street (that you didn't even film while you were on set), it'll be able to intuitively blend and incorporate it into the footage you filmed, filling in additional data with high resolution stills and any additional HD video clips it could find filmed on the same street. Not to mention allowing you to go back after the fact and film more yourself and let that fill in. I think it'll begin to become common practice in the future to do a quick 360 of your area before you start filming a scene, so you'll easily have a scan of the environment that you can easily adjust yourself, and incorporate and change around elements in the computer, treating it like you were actually there on-set adjusting everything.

So if you filmed a scene of someone running down a street in profile, and ONLY filmed that. In post you'll be able to incorporate all of this additional data from cloudspace, have the system make educated guesses about what his other side looked like, then flip around your angle a full 180 degrees.

So you'll be able to adjust focus AND angle in a single video, based mostly on how it incorporates additional data AND how it interprets all of that data.

Also, we'll see as everything is broken down into 3D elements (that can be individually modified, didn't like how high his knees were kicking up as he ran? Adjust his gait in post.), it'll also be able to recognize EXACTLY what the elements are. So it'll be able to think like a human and say, "Ok, that is a car, that is the car's tire, that is a restaurant front, the coat the actor is wearing is leather, etc." That way, even though we'll be inputting massive amounts of resolution data, if something isn't as clear as you'd like, it'll be able to take data about what kind of texture leather has, and fill in additional leather detail, reskinning the leather jacket as a separate 3D element.

Or hey, don't like the jacket? Take it off and replace it with a cotton blazer, and adjust the color.

It'll be able to understand what it's looking at, and adjust it's texture based on what it knows about how that kind of object is textured to fill in and enhance the data it's already capturing if you're pushing it in a direction where it lacks data.

Likewise easy auto-keying, since it'll be able to do depth maps, and break everything down into 3D elements, it'll be able to easily key an actor out of a busy scene, even with the actor having very close colors to the background.

I think it's going to be less about "sensor wars", and more about how the systems can interpret the data. Because within 20-30 years we're going to get a point where the unaided human eye will be able to discern no additional data from the images we're capturing (not saying anything about AIDED human eyes though...), so we'll be capturing the data for easier intuitive breakdown of the objects in the image, not so much to make it clearer for human eyes, but for the system's sake.

Not to mention, lets say you film a scene, I envision being able to "dial in" details that weren't there. Because as it identifies objects in the scene, it'll be able to bring up metadata about them. Like that car in the background as your running? It's a 2014 Toyota-Ford and you'll be able to bring up the exact name of the color, and all the specs about the car, but you'll be able to take all of these elements (like cars in the background), and adjust it's color to whatever you want (Like lets say your colorist wants it a slightly different shade so your actor pops out from the background a little more).

OR, lets say you're doing a period piece, easily identify the manufacture year of all of the objects in the scene and null and remove everything that doesn't fit in with the year it's in. THEN pull up historical photographs of what the street looked like that year, and some models of cars from the same time period, and have the system automatically adjust the background models based on historical photographs, styles of the period, and common car types of the area.

So you'll be able to take a video of you following an actor on a 2030 street, then in camera breakdown all of the objects in the scene, and change them out for models from whatever period you want your camera filming in. So you won't be using so much lens filters in the future as "Reality filters". You'll be able to program a realtime augmented reality for your camera to see.

So period pieces, in real time. I think that'll be one of the most impressive things we'll see in the future.

[Eki Halkka]

So period pieces, in real time. I think that'll be one of the most impressive things we'll see in the future.

Great vision ;-)

We're not really THAT far from a lot of those features... the biggest Epics should have enough resolution and sensor size for good quality plenoptic capture (you just need to add the microlenses). Plenoptic footage has multiple viewpoints in each frame, so with proper 3D tracking / image modeling software, theoretically one should be able to extract 3D data even from a single frame... and so on.

It's the "i understand this is a car" part, or content-awareness, that's probably going to remain the hard challence for programmers even when computing power exponentially grows.

[Wesley Scoggins]

Also, no more Mo-Cap. Tag someone as a Werewolf in your cameras touchscreen, and have a Werewolf model overlayed their body realtime in the camera.

Just as there are Stock photo and video databases now, there will be 3D model databases for every object you can think of. Famous people, landmarks, objects. Easily downloaded, and manipulated by your camera. (On Databases that will more often than not be updated and constantly enhanced by volunteers working for free similar to wikipedia, except it'll be people scanning objects by just walking around them with their cameras, and creating models practically indistinguishable from the real thing.)

You'll also be able to control interactions between these objects in-camera. Tag someone as a bulldozer, and have them run down the street, and have them push and crush objects realistically (as it figures out what the objects are made of and tags them, it'll be able to bring up their physical characteristics such as their density, etc.) as they run down the street.

[Eki Halkka]

Just as there are Stock photo and video databases now, there will be 3D model databases for every object you can think of. Famous people, landmarks, objects.

We already have a beginning of that: there's plenty of stock 3D objects available, and many cities are being modeled in 3D...

...but what will be cool, is that when automatic image based modeling at high quality becomes a reality, we don't need to do that stuff by hand like now: automatic search robots can scour the web for geotagged photos and videos from around the world, and do the modeling automaticly. Like Microsoft's photosynth meets Google earth on steroids, the whole world as a high quality virtual copy.

[Wesley Scoggins]

Exactly. Each year we're seeing the quality of the "streetview" getting better and better (and also getting better at filtering out the irrelevant data, pretty soon it'll be able to combine detail from multiple scans to completely filter out people and cars on the street.) and we'll eventually have a very high quality (and very up-to-date) streetview scans of the majority of the planet that is accessible by roads.

Need your actor to run down a street in New York, and you're in Wisconsin and can't afford the plane tickets fro the whole crew?

Film your guy running down your down town, pull him out of the frame, bring in data from Streetview, incorporate it behind him, conform the colors to your footage, and throw in realistic bokeh, and your guy is in New York.

Throw in some NPC's to do day-to-day normal stuff behind him (no more Extras either), and render out.

Eventually it'll be so good it'll be indistinguishable from filming it for real, and it'll all be in-camera. There will be less distinction between home computer, editing software, and compositing software. It'll almost all be done "in-camera", with the camera pulling computing power from integrated computer systems on your body (everything from the computers in your Smart Shoes that'll have an integrated pedometer and accelerometer, and other integrated systems to determine how good your posture is and how many calories you're burning walking, and other health determination systems, to the ones in your sunglasses with integrated Augmented Reality overviews, to the ones in your wristwatch and Smart Phone system, each individual integrated system will be more powerful than most modern desktop systems.) as needed.

You'll be doing the filming on your camera, have it sent to an internal editing system that'll be able to identify what's happening, and relate that to your script, and start putting your shots in order for you.

Then with a combination of verbal commands and multi-touch interfaces (probably an integrated eye tracking or gesture based control systems as well) be able to quickly do the fine editing (since the system has done most of the heavy lifting for you).

There won't really be a need for separate dedicated systems for each individual step in the future, it'll all be integrated in a single device.

[Mark B.]

For near-term sensor improvements, there's research being done with pink silicon that's apparently close to being in production. It provides improved dynamic range over traditional silicon.

http://www.technologyreview.com/computing/22975/

[Jeff Kilgroe]

There will probably come a day when digital imaging systems no longer acquire or manage image data with "pixels" or the use of individual photosites. But think of a device capable of capturing and recording every photon (and possibly other waves/particles) that it comes into contact with.

[Tom Lowe]

If Jim makes bank off of DSMC, he should set up a Skunkworks/Darpa type secret unit to look into all of this, and place the headquarters building out by Groom Lake and Area 51. :coolgleamA:

Red Ranch is already quite near Nellis Air Force Base and Area 51.

[Raul Gonzo]

Right now, there is a camera that is not a camera, but more like a laser that records everything into 3D space.
I think there is an amazing future in this (http://code.google.com/creative/radiohead/).
http://www.aaronkoblin.com/work/rh/city1.jpg

[Erik Bien]

This sounds like a VFX guy's version of paradise; to do away with old art department/SFX dinosaurs like me ... :001_unsure:

As far as I'm concerned it will be a sad day when we no longer get to travel to far-flung locations, dress people in funny costumes and put them on horseback, burn stuff down and blow stuff up. :crying:

[Raul Gonzo]

I think you would still want to do things like go to the right location and costumes and such. But think of the possibilities of this in post.
You would have a very god-like power to:

Put cameras wherever you want.
Add/subtract lights.
Change time of day.
Move cameras in ways that are nearly impossible in reality.
Endless possibilities...

[Quentin Brown]

Some pretty wild stuff here. But quite possible I'm sure.
The only question on my mind is whether movies will exist in their current format by the time all this becomes reality. Surely things are likeley to progress to a more interactive 3D experience by then. I guess this would be the technology to enable that.

[Quentin Brown]

I'm sure there will always be a place for cinema though;-). Just it's place in the scheme of things will continue to change.

[David Rasberry]

Also, no more Mo-Cap. Tag someone as a Werewolf in your cameras touchscreen, and have a Werewolf model overlayed their body realtime in the camera.

Talk about taking the "skins" concept to a whole new level!:001_cool:

[Eren Ozkural]

I've always imagined the future of cameras as being a cross between LIDAR and plenoptic sensing instruments placed around a "volume" and then just capturing the pure performances and deciding on the style and point of view of the movie later.

I find this sad when I think about it at first but then it dawns on me that non linear editing and CGI would be seen as scary, unpure or fascinating way of making films over a hundred years ago. Almost akin to magic.

[Eren Ozkural]

This talk of the Radiohead video got me thinking about creating the effect at home without expensive scanners.

You just need to replicate their set up and interpret the results through Trapcode Form on After Effects using Luma levels as a displacement map.

I made a quick example video for you guys that's rendering right now but here's some stills to illustrate how a quick and dirty version of this setup looks without polish, error correction etc.

http://farm3.static.flickr.com/2633/3953112230_f37e512a86_o.jpg

http://farm3.static.flickr.com/2581/3953112932_84d547c2d9_o.jpg

[Roberto Lequeux]

Do we have an ETA for my bass with 28k Epics mounted yet?! Dammit Jim!!
How am I supposed to record my world take over without them?!

...could you offer a x4 bundle, please?

[Tim Whitcomb]

I'm not sure Red would have much problem with becoming the next Canon (and then some)


I hope they DON'T become the next Canon, as I believe it's counterintuitive to their mission... obsolescence obsolete is the theme here and why most of us jumped on board.

RED is on a sort of Star Trek... "To boldy go where no one has gone before"

But if it all stopped right now... I'd be happy with my RED ONE and Build 21
for the enxt decade because the images (even transcoded to proress 4444) are just breathtaking...

I'm more worried about the decline of the US dollar than I am what EPIC will look like in the future.

cheers!

[Wesley Scoggins]

http://vimeo.com/6496886

Here to illustrate a prediction I made earlier in the thread. I definitely think we'll start seeing stuff similar to this embedded in cameras and applicable to video within the next 20 years.