Speaking of pulling focus automatically, Philip Bloom took some time to evaluate Canon’s new dual-pixel autofocus on the C100. After a lengthy intro from Bloom about why he prefers to pull focus manually and the challenges of video auto-focus, he finally lets Canon convince him to try it out:
For me there are two types of shooting where if it worked, autofocus could be very useful. Firstly where you couldn’t touch the lens…steadicam/ glidecam/ movi type devices. Secondly, wide open shallow depth of field, where human error isn’t too likely to get good results.
So there we go. I was going to shoot on the gorgeous Movi (a piece of gear I have had for a few months and have gotten quite good at, but mostly have to use in single operator mode which is very limiting due to focus issues. You really need a focus puller!) , some hand held wide open shots and some tripod shots.
And his verdict? Positive, but tepid:
You will understand more once you watch the video below but one of the biggest issues for me was the limitation of the auto focus area just being that small square in the middle. Never in every shot are you going to want your focus to be dead centre.
He adds later on:
Am I a convert? I would say yes…I am impressed enough to be booking my C100 in when I get back from NAB and also my C300 in May. How much will I use it? No idea. I think on the Movi I will use it a fair amount actually. That’s why it’s a shame it won’t be on my Movi camera of choice, 1DC. At least my C300 will have it, and this feature with the Zacuto grip relocator could be the answer to my one man 3 axis brushless gimbal problems!!
Just in this iteration I really wish we could move the area to be measured ourselves, and also change the reaction speed of the autofocus. Fast works well, slow also works well at times!
Going back to my opening paragraph about I don’t believe it will ever work the way I want it to, why is that? It simply cannot read my mind. That is what autofocus needs to do, to understand what I want in the frame to be in focus etc…Once we have that bluetooth implant for the cerebral cortex available, we will have the best of both worlds! Mind reading autofocus with mechanical precision…then again will my mind deliberately make the autofocus no longer work as well? Perhaps it will be trying to replicate the (flawed) human touch…in which case we are back to square one…
Isn’t this the problem with auto-anything? It can never factor the human element into its complex calculations and algorithms. Auto-exposure always shoots for “correct” exposure even though an “incorrect” exposure – by the software’s standards – may be more artistically interesting and suitable for a story. Auto white-balance can get all sorts out of whack if you mix color temperatures. And the auto-ISO function on my DSLR often overestimates how bright I want an image to be.
Auto-focus, too, has these issues when applied to digital cinematography. In Canon’s implementation, you’re limited to focusing on only what’s in the center of the frame. What happens if the chest of talent is in the auto-focus area, but they’re leaning forward and their eyes are not? What if you want to have two objects on each edge of the frame and rack focus between them? And with that rack focus, by the way, you can’t choose the speed of it (which has a strong effect on the mood of the shot).
These auto-focus advances are technically impressive and I’m sure some people will have a use for it, but ultimately it’s limited in function. For now, pulling focus is still the realm of the camera assistant.