Be afraid. Be very afraid

[Aaron Green]

I've seen some demos of cinetape + Arri WCU + focus motors performing "auto focus" using Master Primes.

 

[Satsuki Murashige]

I've seen some demos of cinetape + Arri WCU + focus motors performing "auto focus" using Master Primes.

Yes, but the topic was about how good the camera sensor-level autofocus has become. Cinetape has a lot more limitations, being based on ultrasound. I stopped pulling focus before the Light Ranger 2 came out, so I have no idea if that would be a viable option for autofocus. But I think in order for DPs to accept autofocus more broadly in narrative and commercial settings, it would need to work with their favorite lenses.

 

[AndreeMarkefors]

Useless for scripted work.

That's right up there with "250 GB of hard drive is all you'll ever need in your lifetime" and "the internet is just a fad".

No one wil wait on the set for doodling the camera algorythms to mimic a human hand.

That's the whole point right? To not mimic some poor focus pullers fumbling with constant over/undershoot in a tricky scene, but to actually nail focus (when that's the intention).

It's not either or, it's just a tool.

 

[David Grigorian]

I was at a early screening for the film Uncut Gems where after the film the directors the Safdie brothers spoke about the production of the film, Josh Safdie in particular mentioned how Darius Khondji along with first assist/focus puller Chris Silano used a autofocus/focus assist system called Light Ranger 2 made by Preston. Josh raved about the system and even went as far as saying that without it it would have been impossible to get certain sequences shot, in particular the city exterior street scenes that they filmed on highspeed anamorphic telephoto lenses wide open, they wanted to give the actors freedom to move in the scenes naturally without restrictions or focus marks, they also were filming on live streets with real pedestrian and not a closed set. He said the Light Ranger 2 was a game changer and contributed greatly in achieving the intended look and feel of the film. If you look at demos of the system you can see just how impressive it truly is. My hope is that this type of tech makes its way down into more affordable systems in the future.

https://www.fdtimes.com/2018/01/29/gods-of-focus-speak2/

PS the film was fantastic, i cant wait to see it again when its out in theaters

 

[Björn Benckert]

I can see many shooting situations where this tool would be helpful. But as a cinematographer, the biggest obstacle I see is having to choose between the look of the lens that you want and the utility of the autofocus tool. No matter how good the tech gets, there are many of us who will continue to want the look of Cookes, Zeiss Supers, Master Primes, anamorphics, etc. The lens choice makes a such a big difference to the quality of the final image that I would hate to be confined to shooting only with clinical modern autofocus lenses. If they can also make this work with any lens by outputting the camera’s data through a cable from the lens mount to an external MDR, then I think more cinematographers would choose to use it on bigger budget projects.

There is no reason why pixel based auto focus would not work with external motors.

 

[DJ Meyer]

That's right up there with "250 GB of hard drive is all you'll ever need in your lifetime" and "the internet is just a fad".

Motorized horses!

What is this world coming to...pffft.

 

[Björn Benckert]

Useless for scripted work.

No one wil wait on the set for doodling the camera algorythms to mimic a human hand and perfectly adapt to shot rythm and context according to director's intent.


"Everybody on the set wait 'till I set this thing up".

/money rolling away, actors stuck in emotion on hold/

"Thank you bye bye"

/Enters AC ninja hitting focus without the monitor/


Also, when you tap on the focus point you are already late.
Focus pulling is predictive. Focus puller has to follow action, anticipate and adapt in real time. There is no stuff to program, actors, camera and scene change and improvise on the go.

Eye tracking, cat tracking, whatever tracking is useless if you have a rack focus from eyes to a cat to whatever, exactly at the right moment depending on a/b/c otherwise shot context changes, shot is not editable, shot makes less sense on none at all.

Hmm, strange two big productions here in stockholm just started using patented beta versions of autofocus gears. With DPs and directors with several awards under their belt. They just love it.

Then also, What are you talking about? You say focus pullers just go on intuition? Needs no time to set up? Does not need to mesure thier tape marks? Never ruin a god take on missed focus so there needs to be a new take?

I dont know, I seen alot if tge above. I seen difficult focus work where a few hundred people been standing around waiting for focuspullers to nail thier take and the missed takes counts goes to the roof.

Also done plenty of moco work where it can take days to set up focus targets for a take, where as a keyframe string from a autofocus would be a golden starting point. Run through the take with auto focus, clean the keyframe string, edit whats need to be edited and roll with an enhanced focus keyframe string for the actual takes.

Not to mention, these days they use gimbals, cranes, stedycams etc in scripted work. And yes alot of those times ac’s struggle way more with estimating distance by eye or pull focus of a monitor etc, then they would do with face recognition focus or touch screen driven focusing.

So yes saying its not for scripted work makes me doubt, have you been to alot of such shoots?

 

[Gabriel Mann]

Shockingly Autofocus is a tool and a quickly advancing one for motion these days. It's honestly not needed for much of my work narrative or otherwise, but I 100% see applications where it shows it's advantages.

No need to have a war against it. It's been around for a long time and getting better ever 3-9 months or so.

Preeeeeeeeeecisely. I'm not sure why we are all pretending that every job we do is a tentpole feature film shooting on Master Primes. Jib and steadi ops at music festivals, documentary work, remote-mount applications on narrative work...there are so many applications where that C300M2 kinda AF would be very, very, very clutch. No need to flex.

 

[Ted Parkes]

I have been shooting for 30 + years and I for one love the advancements in Auto Focus. I have seen the demos with the Sony FX9 and was extremely impressed. This won't make any shot obsolete but it may make an impossible shot possible. Bring it on!

 

[Hrvoje Simic]

That's right up there with "250 GB of hard drive is all you'll ever need in your lifetime" and "the internet is just a fad".

For you currently.
Because you draw innacurate analogy through superficial interpretation of some kind of stupid resistance to technology.

In reality - not even close.

That's the whole point right? To not mimic some poor focus pullers fumbling with constant over/undershoot in a tricky scene, but to actually nail focus (when that's the intention).

"Actually nailing focus" is mostly achieved already.
There is much more to it than just nailing focus.

It's not either or, it's just a tool.

Exactly.
And tools have their place.

 

[Hrvoje Simic]

Hmm, strange two big productions here in stockholm just started using patented beta versions of autofocus gears. With DPs and directors with several awards under their belt. They just love it.

Not strange at all.

Nor is the techy infatuation and depreciation of human work.

Then also, What are you talking about? You say focus pullers just go on intuition? Needs no time to set up? Does not need to mesure thier tape marks? Never ruin a god take on missed focus so there needs to be a new take?

You seem to have trouble with logical process again, and the scenario is similar as the last time, including a) technology and b) someone else's work, where the outcome is the same > depreciation of someone's work (the one you are not doing) through infatuation with tech.

Last time it was the colourists.

I can help in pointing the error in reasoning if necessary.

I dont know, I seen alot if tge above. I seen difficult focus work where a few hundred people been standing around waiting for focuspullers to nail thier take and the missed takes counts goes to the roof.

Yes, this is one of the errors in reasoning.

And yes, I've seen it too.
And still stated what I did.

Also done plenty of moco work where it can take days to set up focus targets for a take, where as a keyframe string from a autofocus would be a golden starting point. Run through the take with auto focus, clean the keyframe string, edit whats need to be edited and roll with an enhanced focus keyframe string for the actual takes.

Then you use robots where that makes sense and use a Human when it doesn't. In the context I mentioned it mostly doesn't.

So yes saying its not for scripted work makes me doubt, have you been to alot of such shoots?

Not for the last month.

 

[Mike P.]

Motorized horses!

What is this world coming to...pffft.

Off topic, but someone please tell me I'm not the only one that reads DJ's posts in Randy Marsh's voice.

 

[John David Pope]

Also, when you tap on the focus point you are already late.
Focus pulling is predictive. Focus puller has to follow action, anticipate and adapt in real time. There is no stuff to program, actors, camera and scene change and improvise on the go.

But it doesn't have to be all AF or all manual. If the focus needs to be predictive, the focus operator takes over and makes that happen. But in the meantime, for "just" holding focus on a subject, the AF can do that and in many instances do it better than can a human. Sort of how cruise-control on a vehicle maintains until the operator, the driver, chooses to take control. The fact that cruise-control can't anticipate a need to reduce speed or brake doesn't render cruise-control useless.

I don't see where there would be any messing with algorithms on set. ( besides, with human focus pulling there is down-time when the focus puller's Preston or Light Ranger has an issue. ) The operator simply touches on screen what he wants the AF system to keep in focus. The actors, the rest of the crew, would never even know when the 1st AC is pulling manually or using AF. Just as the rest of the crew never knows if the 1st AC is pulling by eye alone or pulling to the Light Ranger indicators. Or some combination of both.

I do agree that where no AF system can be of use is when focusing to a point in space. In other words, you need the focus to be on a point where something is going to appear but has not yet appeared. That said, with human focus control there can be inaccuracy in such situations when the thing appearing misses the mark and the human focus puller has to adjust on the fly to that miss.

" Focus puller has to follow action, anticipate and adapt in real time. There is no stuff to program"

Is that true? Don't focus pullers program their Preston's and make adjustment to the programming of their Preston and Light Ranger during the shooting?

 

[Björn Benckert]

Not strange at all.

Nor is the techy infatuation and depreciation of human work.





You seem to have trouble with logical process again, and the scenario is similar as the last time, including a) technology and b) someone else's work, where the outcome is the same > depreciation of someone's work (the one you are not doing) through infatuation with tech.

Last time it was the colourists.

I can help in pointing the error in reasoning if necessary.



Yes, this is one of the errors in reasoning.

And yes, I've seen it too.
And still stated what I did.




Then you use robots where that makes sense and use a Human when it doesn't. In the context I mentioned it mostly doesn't.




Not for the last month.

You really got some built up aggravation against me. Sorry for making you so edgy.

But I'm absolutely not here to try "depreciation someone´s work".

I'm just stating that in the future I think focus pulling will not be so much about putting tape marks and using a tape measure as it has been it will be a way more computerized and digital artform. Kind of the same evolution as for like second camera assists who does not deal so much with loading celouid film rolls any more. Its just me reflecting over the digitalization of this industry. So I dont mean to piss at anyone when I say, AC work will be different in the future.

And yes, a while ago dropped some comments about colorists, I think I was saying something about that the industry is changing and a huge chunk of grading work will be done by none full time colorists, something like that. And that I found problems working with colorist that did not want to keep an open timeline / do the grade in resolve and leave it open for VFX. Well I now a few years later so a lot of people around here work like that. I also see the resolve user base increase in a way so I'm sure its not only full time colorist using the tool. But I remember it really pissed you and a few other colorists of here. Did not mean to upset you guys, it was just then my reflection of where I see things are heading.

So please Horjve drop it, I do not mean to de credit colorist, focus puller or what ever other professionals in this industry. All I'm saying is the artform of making films is constantly changing and autofocus I think is the next big change of the game, and I'm sure it will happen very soon as its one bit of filmaking left to digitize and the key solutions are now already there, sensors and camera processors and IP to do this is now available, people is already using it. Not for every possible scenario of course but for a lot fo things, the number of way so apply it will grow when this tech get more sorted.

Preferred type of lenses will change, Im sure fast internal focus motors will be more of a key function. Data passing from the lens will also be more key. Metadata per frame for lens data / focusing and zoom values etc will also dictate future choise of glas. The more data the captured file can tell about the lens the more control for simple and accurate post enhancement of the captured material. Its very possible for a lens / camera manufacturer already today create OFX plugins for their lenses. As the meta data hold focus, zoom and sensor size values is a no brainer to render the captured material to a perfectly rectinear none breathing les vingeting image with reduced cromatic aboration for example. But that does not mean VFX people will not work it just means their work will be more digitized / different in the future.

So yes, I really do not mean piss you off. This is just me speculating around the what will happen in the near future.

 

[Tom Gleeson]

Fascinating how many people see the world in such a binary way. I don't believe focus pullers will be out of a job anytime soon and these AF systems have a way to go yet before they're used on the majority of sets but is is coming. Focus pullers work will be eroded by these systems but unlikely to be replaced. John David Pope's post says it very well.

 

[Hrvoje Simic]

But it doesn't have to be all AF or all manual. If the focus needs to be predictive, the focus operator takes over and makes that happen.

Which needs fiddling and introduces new variables to juggle and a need to rely on an algorythm for a specific task. Instead of manual work. Many prefer the latter. Having firm basis for the preference.

Again, I am talking about typical scripted production set and most of the scenarios in that context. Not about specific cases or the whole spectrum of production. You can do whatever you please when you don't bog down the whole production with fiddling and you bet your own money on algorythms, and if the route helps, great.

And yes, if a shot cannot be achieved by a skilled focus puller (which is extremely rare) it makes sense to use a route which helps to do it.

That is unless here we are just talking about "I want to do this CU tracking shot on telephoto wide open just because I have this idea, without any specific reason or budget for AC/focus puller able to pull it off." Which has nothing to do with serious production.

But in the meantime, for "just" holding focus on a subject, the AF can do that and in many instances do it better than can a human.

Yes it can work.
And in many instances it cannot and Human can do it better.

Sort of how cruise-control on a vehicle maintains until the operator, the driver, chooses to take control. The fact that cruise-control can't anticipate a need to reduce speed or brake doesn't render cruise-control useless.

Cruise control is a solution for basic tasks and adapting to machinery.

Not for actors missing marks, cam ops improvising camera movements, shots which differ while telling a story by presenting meaning based on subtleties and directors with precise requests, changing their minds or wanting variant a/b/c.

Also, focus pulling in the context I am talking about is a creative decision and judgement call type of work, not just precision and mechanics.
Which needs a Human.

I don't see where there would be any messing with algorithms on set.

Where you'd constantly fiddle attempting to optimize the algae behavior to what I mentioned above. Which changes constantly and in the moment.

That said, with human focus control there can be inaccuracy in such situations when the thing appearing misses the mark and the human focus puller has to adjust on the fly to that miss.

Humans also make autofocus tools and may miss there as well.
Autofocus doesn't know which part of the actor has to be in focus and when.

Btw focus puller's job is just that, adjusting on the fly. Knowing the context, and just the right feel the director wants. And they do it very well.

They just don't have the hype so many are not aware of the skills which make their favourite films or series happen and not appreciate the collaborative effort of many, which gives the Soul to everything.

The multitude of subtleties, Human touch, the finesse, is what makes the work special.
Generic stuff....doesn't.

Don't focus pullers program their Preston's and make adjustment to the programming of their Preston and Light Ranger during the shooting?

Those tools are not replacing typical focus pulling, they expand insight and options.

 

[Karim D. Ghantous]

Eye tracking, cat tracking, whatever tracking is useless if you have a rack focus from eyes to a cat to whatever, exactly at the right moment depending on a/b/c otherwise shot context changes, shot is not editable, shot makes less sense on none at all.

I disagree with this perspective. Take Sony's A9, for example. It is amazingly accurate and precise. A tennis player can be moving back and forth, behind the net, whatever, and the A9 locks onto them like a rabid dog on a baby. AF algorithms can certainly work while a frame isn't being recorded (e.g. the 180 degrees that the 'shutter' is 'closed').

If I make serious money with my crypto investments, I'm going to start an optics company that solves the documentary/ENG focus issue from a different angle.

 

[Jaime Vallés]

But it doesn't have to be all AF or all manual. If the focus needs to be predictive, the focus operator takes over and makes that happen. But in the meantime, for "just" holding focus on a subject, the AF can do that and in many instances do it better than can a human. Sort of how cruise-control on a vehicle maintains until the operator, the driver, chooses to take control. The fact that cruise-control can't anticipate a need to reduce speed or brake doesn't render cruise-control useless.

I don't see where there would be any messing with algorithms on set. ( besides, with human focus pulling there is down-time when the focus puller's Preston or Light Ranger has an issue. ) The operator simply touches on screen what he wants the AF system to keep in focus. The actors, the rest of the crew, would never even know when the 1st AC is pulling manually or using AF. Just as the rest of the crew never knows if the 1st AC is pulling by eye alone or pulling to the Light Ranger indicators. Or some combination of both.

I do agree that where no AF system can be of use is when focusing to a point in space. In other words, you need the focus to be on a point where something is going to appear but has not yet appeared. That said, with human focus control there can be inaccuracy in such situations when the thing appearing misses the mark and the human focus puller has to adjust on the fly to that miss.

" Focus puller has to follow action, anticipate and adapt in real time. There is no stuff to program"

Is that true? Don't focus pullers program their Preston's and make adjustment to the programming of their Preston and Light Ranger during the shooting?

This.

I'm at the point where I won't buy any more lenses unless they have built-in auto-focus motors. It's just too useful not to have it.

 

[Hrvoje Simic]

You really got some built up aggravation against me. Sorry for making you so edgy.

There is zero aggrivation and edgyness or any negative emotion. It just seems that way because of the critique.

This is not a personal matter, you and me are irrelevant here, perception of the filmmaking craft isn't. Some things have to be stressed to keep the socially inherited latent technophilia under control. I am aware of the harshness, it is due to seriousness of the matter. This topic is not new and pops out every now and then.

I'm just stating that in the future I think focus pulling will not be so much about putting tape marks and using a tape measure as it has been it will be a way more computerized and digital artform.

Everything gets more computerized by inertia and if we are not careful it may lead to getting lost in handling computers instead of matters directly, deterioration of human skills, dehumanization of the work, and generic souless crap. It is already happening in other domains. Let's keep the awareness and choose wisely what we appreciate and celebrate and who is the master and who the servant.

That's what I am saying.

Btw, I rarely see actual tape now on set, mostly laser "digital tape" measure to check, tape is sometimes there for hands reach close delicate shots. Which also matters less with actor/camera variables and improvization. Generally, AC has his follow focus rings prepared on camera test day with markings set per lens, and measures on set just to check here and there. Most stuff is in the head and muscle memory already with distance judging ability and eye-hand coordination trained to the level of excellence.

This is normal procedure used every day and generally there is zero desire to complicate this process. No need to fix what is not broken. Minor tweaks and aids, sure. Completely changing the workflow principle is unlikely.

Kind of the same evolution as for like second camera assists who does not deal so much with loading celouid film rolls any more. Its just me reflecting over the digitalization of this industry. So I dont mean to piss at anyone when I say, AC work will be different in the future.

I was not suggesting you did anything on purpose. It is already different in some ways but some principles are mandatory, which include direct hands on, speed and practicality, where tech in between can many times act as limitation, not as aid.

Evolution is not just addition or complication or technological, evolution is also reduction, optimization to purpose and mental. Evolution are skills, not incompetence pampered with complete dependency on computers.
Tool practicality, intuitivity, speed, reliability, adaptability, gear setup, gear maintenance, camera prepartion time, setup change time, location change time etc etc. Skilled AC makes all the difference and tech setup has to be optimal, not drive-by wire indirect tech overkill.

Preferred type of lenses will change, Im sure fast internal focus motors will be more of a key function. Data passing from the lens will also be more key. Metadata per frame for lens data / focusing and zoom values etc will also dictate future choise of glas. The more data the captured file can tell about the lens the more control for simple and accurate post enhancement of the captured material. Its very possible for a lens / camera manufacturer already today create OFX plugins for their lenses. As the meta data hold focus, zoom and sensor size values is a no brainer to render the captured material to a perfectly rectinear none breathing les vingeting image with reduced cromatic aboration for example. But that does not mean VFX people will not work it just means their work will be more digitized / different in the future.

It is already highly digitalized. We already pass tons of data to post, some of which is automatic, some of which we have to input manually for each shot. For each camera and each shooting day. Tons of info and reports to handle, check and re-check and carry the responsibility. One can simplify and make more practical the information gathering and post workflow implementation, that is reasonable.
Re-shaping the production workfow is another matter.


Example: in theory the depth sensing system could map the space and recognize and track objects in the scene and focus puller could manually rack focus from mappped points of reference, which would keep the manual control and help in repeatability.

But that brings another set of problems where
a) you distribute the responsibility of delicate matters to a machine
b) have to adapt to and pamper the system for achieving something done by humans every day
c) complicate the process of something which in principle is very simple

So in that option...you would change the established workflow which works in 99% of the time...why exactly ? To cover the 1 % ? Do you have time, money and opportunity to go through the phase which has to assure the other 99% is covered as well ? Many productions do not.

So the whole point of many productions it to keep things branching throughout the crew under manual control and each and every person doing his work, able to adapt and improvise instantly. That's why there is camera operator, 1st AC, 2nd AC etc per single camera, not just a camera operator doing everything camera related.

 

[Björn Benckert]

Yes I just see it differently, I see DP´s move to higher Fstops than they first wanted just because its simply not possible to keep good focus for what ever is happening in the scene. I cut film with sissors when I was a kid thats an art form. But I don´t feel the computers killed the art or creativity of what I do in my profession today on the opposite. Same for focus pullers, I think their line of work will get way more interesting in the future so I dont see any of this as a bad thing or work fretting for anyone, but it will for sure race the bar for whats possible. And yes, doing difficult stuff, even technical such is by some also considered an art from.