Need to obscure a face in 100's of hours of interview vids

[IamMyth12]

HI. 

I have a project in which I need to obscure the face of an interview subject over the course of 100's of hours of video.  Preferably, I would like to blur the face, but I've found that using the tracking features in Premiere, or moving around the blur manually, is taking me about 5 hours of work per 1 hour of video.  This is unrealistic, given how much material I have to cover.  I was wondering if there was another solution.  I've seen real time object tracking software on the internet (not sure if this would work, or if it is somehow interface-able with the blurring feature in Adobe's programs).  I've also contemplated the idea of moving an object around with my mouse (like a black circle) in real time, and having a screen capture video program record my movements as I perform them in real time over the face in the original video (not sure if this is possible either - I'm just brainstorming). 

Any ideas or advice would be greatly appreciated.

Thanks in advance.

[Abambo]

Discussion successfully moved from Adobe Creative Cloud to Premiere Pro CC

[chrisw44157881]

perhaps mocha pro plugin for premiere. it uses planar tracking so may be less work.

[TaranVH]

Interesting problem...

But why were hundreds of hours of footage shot of this person, with apparently no plan for how to censor his/her face?

Sounds like a shoot first, ask questions later situation.

If they were so lazy while shooting it, then you can be lazy while editing it.

Make your blur mask larger. Much larger. Have it cover all the area where this person will be shifting around. (but beware of feathering - don't let your subject even touch it.)

So, instead of this:

https://i.imgur.com/wBGbZQC.jpg

do this:

https://i.imgur.com/nVaR7ee.jpg

Also, why are you having to censor "hundreds of hours?"

What is this, like, a documentary?

Just cut the video together, get picture lock, and THEN censor the guy's face.

There's no point wasting time censoring a face in footage that might not even make it to the final video.

[Abambo]

TaranVH  wrote

There's no point wasting time censoring a face in footage that might not even make it to the final video.

I suppose this is the solution.

And if there is a programme of 100 hours of footage... then there need to be the time to invest to do the work that is needed. But probably someone took on the editing job to edit 100hours in 2 days...

[IamMyth12]

I appreciate the response. 

It is not,  in fact, a documentary.  More like a video installation/performance art type piece.  All of the video will be used, so I can't just cut together a shorter, final version.  I have time to get it accomplished (I did not agree to edit 100+ hours of video in 2 days), but I don't have thousands of hours to dedicate to the project, nor is there the budget for me to do such a thing. 

This brings me back to the original question.  If there isn't a pragmatic way to do it in Premiere (not sure if the plugin mentioned above would make a significant difference, but am willing to try), what do you all think of the solution of recording the screen and somehow turning my cursor into a black circle and then just manually tracking over the face with my mouse?  Is such a thing even possible?

Making the blur huge is not the solution I would go to if it can be avoided.

[chrisw44157881]

that's pretty ingenious. after effects has a motion sketch effect where it records your mouse in realtime.

[IamMyth12]

Thanks for the encouraging words and the information.

[TaranVH]

You can try to use motion sketch, but I don't think it will work. All layers become wireframes when you use it. you won't be able to see the footage: CreativeCOW

Option 1: Screen recording while your cursor is a big black circle... WILL work.

And yes, it is possible to change your cursor into something else.

The only way I can think of (there may be others) is to use AutoHotKey.

Here is a script that I use for click visualization, that you can use.

You'd have to modify the script to, first, get rid of the scroll wheel acceleration.

The, make it so that instead of a "(" , the visual is a :black_circle: or ▮ or ⬮. And, increase the font size until it is large enough to obscure the person's face (but even bigger than that, to give yourself some wiggle room.)

Find and implement a script to make the cursor invisible.

you'll probably want a toggle key to swap between ⬮ and the normal cursor.

(Don't know how to use autohotkey? Start here: https://autohotkey.com/docs/Tutorial.htm )

You can alternatively use this guy's script as a base instead of mine - it might be easier: https://autohotkey.com/board/topic/32608-changing-the-system-cursor/

Do your screen recording in full screen... make sure the timecode is visible, (you'll crop it out later) using OBS or shadowplay, with good settings that will not degrade the footage. RECORD AUDIO FROM A MIC, (and preferably, record desktop audio as well, but on a seperate track.) so that any time you might fail to move the "cursor" in time and the person becomes visible, all you need to do it shout "FAIL" into the mic. Keep going with your recording. Then, later, when you bring that whole video file into premiere, you can instantly see all the little places where you messed up, and need to re-record those segments.

Now, do that for "hundreds of hours of video."

-------------------------OR----------------------------

Option 2: Just make a huge blob of censorship, everywhere that this guy will move his head, like I showed you, and be done with it.

It will only fail if the guy leans out of his chair or something. you can fix those rare instances manually. but it still means you have to watch the whole d**n thing. but you can do so at double or even triple speed.

I'd go with option 2.  Don't make this much work for yourself. It is NOT worth it.

[IamMyth12]

I appreciate all this helpful information and advice.  I'm not really familiar with any of the processes that you suggest here, but will look over the materials and links you've provided, and see if I can make sense of it.

Things were not done differently from the start not because of a lack of planning, but because this was the desired effect from the get-go.  We just had no idea how much work was implied in its execution. The only option we discussed from the beginning (and it turned out to be way over our heads), was if we could do something similar to what was done years ago in hockey games, where they somehow built a hockey puck that glowed in real time during the live telecasts.  Our idea was to somehow use a mask that would allow for realtime tracking and blurring.  Needless to say, our search for how to do such a thing was fruitless.

Again, thanks for your time and for your generosity in sharing your knowledge.

Cheers

[TaranVH · Accepted solution]

If your search for easy, real-time tracking was fruitless, then you should not have shot the video with the assumption that you'd be able to figure it out later.

It was a lack of planning. You guys should have done a small-scale test. Have a dude talk on-camera for 10 minutes, not 100 hours. Test various methods for obscuring his face as he moves around. See how they fare. Only when you find something that works fast enough, should you decide to shoot the video in that way.

Sound like you guys, (or, perhaps just you in particular) learned your lesson on this one. You bit off more than you could chew, but now you've gotta chew it anyway. Best of luck.

[Abambo]

IamMyth12  wrote

The only option we discussed from the beginning (and it turned out to be way over our heads), was if we could do something similar to what was done years ago in hockey games, where they somehow built a hockey puck that glowed in real time during the live telecasts.  Our idea was to somehow use a mask that would allow for realtime tracking and blurring.  Needless to say, our search for how to do such a thing was fruitless.

My guess is that the hockey puck has been prepared accordingly and that the broadcaster had the budget to invest in the correct software and hardware. I’m sure, you could do what you intend to do only by putting the $$$ on the table.

We are talking, however, about professionels at the low end of the budgeted systems. Even that  a lot happened during the last years, I do not think having such soffisticated software (supported by standard GPU) in the very near future for a couple of $.

[Meg The Dog]

You could do it in real time by playing the file back through a video switcher with an effects bank, and manually tracking the obscuring section with a joystick. You would record the output of the switcher on a mother device.

But this is not going to be faster than real time, if you 100+ hours of source, you'll have to spend at least 100+ hours transferring and obscuring the area you want to obscure. And paying for the use of the facilities.

MtD

[TaranVH]

What hurts is that with the tiniest bit of planning, this all could have been avoided:

- Shoot the guy in silhouette with no light on his face (and then crush the blacks so that no detail remains) (though, I don't like this method - you can still kinda see what the person looks like, especially if they rotate their head)

- Or, just have the person wear a mask for the interview.