A working Scarlet!

mcgeedigital said:
I have several.

That isn't the point.

The point is that defining two touch screen points on a 5" monitor to pull focus on isn't going to be as precise or as predictive as a trained AC with experience and skill.

Period.
Click to expand...

Sorry, but this doesn't make sense. I understand this is how many people have done it in the past and are worried about computers taking over jobs... but there is no way a person can pull focus better than a computer. The auto-focusing algorithms on modern DSLR are so fast an precise it's mind blowing. If the Red's focus-tracking is anywhere near that precision, but for video... and in real-time, everyone can kiss the focus puller goodbye. When you factor in consistency, reliability, fatigue, ect... there is just no why a person could do it better. For a practical example... What if you are shooting a VFX plate that requires two passes with an identical focus pull? The only option would be programmable and repeatable focus. The only situation I could see a person being beneficial is if the focus-tracking is miss-programed or something and focus needs to be pulled on the fly. But then again... if you can just lock in a subject for continual tracking, then there's no point for a person again. I know it's going to be hard for allot of people currently in the industry to accept and embrace a new technology that takes away yet another job... but, be honest, with all the way technology is progressing with digital cameras, do people really think that focus ISN'T going to be done by the camera in the next year or two?

BTW, the focus is not being tracked by the information from a 5" screen. That is only the input. I'm sure the actual focusing algorithm is being done internally on a pixel-to-pixel scale.
 
Cbrianchester said:
Sorry, but this doesn't make sense. I understand this is how many people have done it in the past and are worried about computers taking over jobs... but there is no way a person can pull focus better than a computer. The auto-focusing algorithms on modern DSLR are so fast an precise it's mind blowing. If the Red's focus-tracking is anywhere near that precision, but for video... and in real-time, everyone can kiss the focus puller goodbye. When you factor in consistency, reliability, fatigue, ect... there is just no why a person could do it better. For a practical example... What if you are shooting a VFX plate that requires two passes with an identical focus pull? The only option would be programmable and repeatable focus. The only situation I could see a person being beneficial is if the focus-tracking is miss-programed or something and focus needs to be pulled on the fly. But then again... if you can just lock in a subject for continual tracking, then there's no point for a person again. I know it's going to be hard for allot of people currently in the industry to accept and embrace a new technology that takes away yet another job... but, be honest, with all the way technology is progressing with digital cameras, do people really think that focus ISN'T going to be done by the camera in the next year or two?

BTW, the focus is not being tracked by the information from a 5" screen. That is only the input. I'm sure the actual focusing algorithm is being done internally on a pixel-to-pixel scale.
Click to expand...

This statement made me come out of hiding. Haha.

I would wager that focus will not be done by the camera in the next ten years. An optical system will not be capable of predicting the movement of a human like a seasoned camera operator and focus puller will. Not any time soon, no way.

Pulling focus is more than just pointing it at something and letting it focus... in cinema, it's anticipating every little head nod, turn, tilt, shift, bend, bump, jump... should I continue?

In addition to that, it'll take much more time to make the focus pull smooth and less mechanical.

This is all to say I love technology too, but I highly doubt any auto-focus is going to be able to compete with a real human being for quite some time where narrative motion picture and similar is concerned.
 
Kholi said:
This statement made me come out of hiding. Haha.

I would wager that focus will not be done by the camera in the next ten years. An optical system will not be capable of predicting the movement of a human like a seasoned camera operator and focus puller will. Not any time soon, no way.

Pulling focus is more than just pointing it at something and letting it focus... in cinema, it's anticipating every little head nod, turn, tilt, shift, bend, bump, jump... should I continue?
Click to expand...

Really? How shallow are you talking? 5D at f1.2 or Imax? I don't think with 2/3 chips or even s35 a little head nod is going to throw the focus off. But even if it did... I assume the focus-tracking for video would be running at double the frame-rate. So it wouldn't really have a need to react to movement, it would just be there already. Although, I don't know how thier system works, or if it's even capable of just locking on to a face and fallowing it. But this is a good discussion still. I haven't seen much of human vs. computer FF comparison.
 
Ok... How hard is for a company to make a device that you stick on an actor and it pinpoints its distance according to the camera focal plane and feeds it to the lens for automatic adjustment?

But I'm with Mcgee, sometimes focus is more than just following object X, you need to change, adapt, compensate for an actor breathing when you're at T1.3 at close focus, I agree a focus puller will always be needed however...

I was just curious about Scarlet digital FF capabilties, which is a nice addition to the low-ish budget crowd of Scarlet.
 
Pietro Impagliazzo said:
Ok... How hard is for a company to make a device that you stick on an actor and it pinpoints its distance according to the camera focal plane and feeds it to the lens for automatic adjustment?
Click to expand...

It is not that hard.
You need small device with mike and ultrasound sound generator on micro battery.
Same on camera.
You'll need to send ultrasound signal.
Device will sence it and send signal on other frequency back with small predeterminated delay.
If you know temperature, you could calculate distance very accurately :-)
 
Pulling focus is easy for a computer when the object is in the center of a frame and stays there and/or choreographed, but there are so many other instances where the actor is moving through the frame in unpredictable ways amongst other actors that human direction of a FF is a necessity. If everything could be programmed, including the actor it would be a piece of cake. But unlike computers, actors are irrational beings and proud of it. In fact that is what makes them actors.
 
Pietro Impagliazzo said:
Ok... How hard is for a company to make a device that you stick on an actor and it pinpoints its distance according to the camera focal plane and feeds it to the lens for automatic adjustment?

But I'm with Mcgee, sometimes focus is more than just following object X, you need to change, adapt, compensate for an actor breathing when you're at T1.3 at close focus, I agree a focus puller will always be needed however...

I was just curious about Scarlet digital FF capabilties, which is a nice addition to the low-ish budget crowd of Scarlet.
Click to expand...

Well, as to your first question, it's been done somewhat. Google cinetape or panatape.

But the real problem with autofocus isn't following "an actor". It's following the ACTORS in a scene. Because unless all you do is focus on one person per shot, the camera cannot read your mind and know that it needs to change focus between person A, person C, back to person A, then over to person C at the back of the room. All in the span of a couple lines of dialog. But the First AC knows that, knows the marks for all of them, and can hit them. While computers may have the speed to out-process a human, they lack the ability to predict. Perhaps at some point when the AF systems can be fed the dialog, and can be fed focus points along with each line of dialog, you might have something. But that's just not possible yet.
 
Barry_Green said:
EX3 doesn't use any manner of Bayer patter, it's a 3-chip camera and that technology is capable of resolving a full 1080p.
Click to expand...
The "technology is capable" on a theoretical level, but not practically. Yes, there is R, G, and B data for each pixel (off the sensors, compression's another story), and each chip has 1920x1080 photosites, but that does not mean it resolves 1080. Forgetting the lens and the prism, I'd imagine the EX cameras have an OLPF, softening the image to avoid aliasing. And that's a good thing - we don't want cameras resolving their full pixel count. It'd look terrible.

I'm not directing this at you specifically, Barry - just seems that when the resolution debates surface, people get on about Bayer vs. Stripe vs. 3-Chip, and forget that none of these cameras are resolving what they "say" they are - fact of life, necessary evil. The best way to deal with this is oversampling - acquiring at a higher resolution than delivery.

But yeah, side note - your numbers are actually a bit outdated regarding RED debayering/resolution. 4.5K mode has been measured at a full 4K, and I believe the most recent measuring of 4K was something like 3.7K. So yeah, you're completely correct in sentiment - I'm just a stickler for precision (sorry about that).
 
mcgeedigital said:
I have several.

That isn't the point.

The point is that defining two touch screen points on a 5" monitor to pull focus on isn't going to be as precise or as predictive as a trained AC with experience and skill.

Period.
Click to expand...

This is what I find interesting about this camera. Hopefully it will focus differently from DSLR lenses. We all know that even the best DSLR lens autofocus isn't up to the task for cinematography, maybe okay for a bit of documentary shooting in a sticky situation, but that's about it. If they can overcome the inherent problem in autofocus, namely the way in which focus changes occur, for example going past the point before coming back, then it might be a really useable feature. But I agree, it can't be intelligent enough to know when something changes in a planned shot and this is where a trained operator would always be more preferable.
 
There is a lot of confusion between output resolution and a sensor's pixel resolution. Red uses sensor's pixel resolution because it outputs to raw, and because its a great marketing tool. 3k, 4k, and 5k sounds good.

The F35's sensor resolution is 5760x2180, but Sony isn't selling it as a 5k sensor. Its not a bayer and it outputs to 2k, so its a 2k camera. Arri Alexa is a 3.5k sensor, but its also a 2k camera.

Its going to get even more confusing when the 4k cameras start arriving. Sony 4k camera is suppose to have a 6k x 3k sensor Q67 rotated bayer sensor, JVC showed their 4k camera to be 3840×2048 senor - but it'll have 3 CMOS sensors.

So there is an apples to oranges relationship in these definitions when you define a camera by its sensor resolution, and especially when its bayer filtered sensor.
 
PerroneFord said:
But the real problem with autofocus isn't following "an actor". It's following the ACTORS in a scene. Because unless all you do is focus on one person per shot, the camera cannot read your mind and know that it needs to change focus between person A, person C, back to person A, then over to person C at the back of the room.
Click to expand...

Maybe I misunderstand this, but I don´t think we are talking about a DSLR "Autofocus" function here.

I think you still need a human focus puller that tells the camera what should be in focus.
But instead of a traditional FF wheel, he uses a screen and his finger to point at the area that should be in focus. So it is just a different input device, everything else hasn't changed.
 
As a director, here is how I would like to see touch screen focus pulling implemented

- an onscreen control to set the speed of the pull
- another onscreen control to set easing for the start and end of the pull
- rack focus is controlled by touching one point then the other at the desired time
- 2 or 3 sided reposition-able sunhood designed similar to smallHD's
- focus assist mode where an area is enlarged and cropped

after all it is 2010, I'm sure we can move on from a plastic wheel and whip
 
Last edited:
- an onscreen control to set the speed of the pull
- another onscreen control to set easing for the start and end of the pull
Click to expand...

If these preferences could be modified and loaded like a scene file, then this would make it even more flexible. Focus pulling might look like this in a few years..

RR002308.jpg
 
I would be agains something so rigid as a scene file. An actors performance changes from take to take.
New things come up and the pacing of the shot changes from the original blocking. Having a human element is very necessary in focus pulling.
My suggestions for having easing and speed were made as they would mimic the operators control over the wheel.
Even better would be a realtime jog rocker to allow the operator to speed up the rack focus part way thru the move if the current
speed setting was slow.
 
Yeah agreed, but I was thinking of easing and out response curves, like adjustable bezier curves that could be loaded with the operators preferences.
I'm sure the Red team have been through the different ways of doing this. Let's hope it is a flexible system. One thing is for sure, if the lens on the fixed isn't up to scratch, then the camera is dead in the starting gates.
 
AdrianF said:
One thing is for sure, if the lens on the fixed isn't up to scratch, then the camera is dead in the starting gates.
Click to expand...

Its a cine zoom which puts it in a class totally above any still lenses you would see made for micro 4/3, nikon, canon, zeiss etc.

For example here is the cheapest comparable lens I found in a quick search.
It's a canon cine zoom covering same image circle and of similar focal length range and T stop, check out the price for lens only
http://www.abelcine.com/store/Canon-6.6-66mm-Super16-Zoom/

note: the close focal length of the canon cine is 60cm
and the close focus of the fixed is pretty much macro link
 
Last edited:
J Davis said:
Its a cine zoom which puts it in a class totally above any still lenses you would see made for micro 4/3, nikon, canon, zeiss etc.

For example here is the cheapest comparable lens I found in a quick search.
It's a canon cine zoom covering same image circle and of similar focal length range and T stop, check out the price for lens only
http://www.abelcine.com/store/Canon-6.6-66mm-Super16-Zoom/
Click to expand...

Here is a 20 times 2/3 inch zoom f1.8-2 for $2000 http://www.bhphotovideo.com/c/product/231972-REG/Fujinon_A20X8_6BRM_A20x8_6BRM_2_3_20_1_ENG_EFP.html
I feel that the Scarlet does offer some interesting design features but I have a problem with a "Pro camera" with a single ring (one ring for all) fixed lens. I think Red way be having a problem deciding on what the market is for the fixed lens Scarlet, it started out as a Pocket Camera then a Pro camera. Since it was first announced the market has changed greatly as we see the price point for drop for flexible 2/3 to s35 provideo cameras. I am sure that Red and DSLR had a lot to do with this change. Will the Scarlet still be the affordable small Pro camera that is better that anything in its price range as it was once touted to be? We will have to wait and see, at least it looks like it may be here in 2011. What a year 2011 will be for new cameras Sony, Panasonic and Red all with new ground breaking cameras to choose from.
 
i dont care till is has a price an is ready to order. all this hype i got over it. its either nut up or shut up already. or GTFO
 
You must log in or register to reply here.
Back
Top