[Mark Suszko]

The green screen is only part of the approach; it alone won’t give you the perspective warps, only a clean place to put the replacement video, which may look weird if the apparent angles don’t match up. And spill or reflection on the set from the green onto surrounding objects can make the comp more difficult.

My preferred approach would be to not put green on the screen at all, but either a flickering blue-white, or nothing at all, just the black screen and natural reflection of the glass. Then use the distort tool and feathered masks to corner-pin your new video into the screen and warp it to match. Then you change the blending mode and opacity so the glass reflection comes thru in the new video, and that sells that it’s coming out of the TV instead of being pasted-on.

Why I would use the randomly-flickering white or bluish-white as an on-set “practical” effect is that, once the screen gets covered, the footage under the mask will still throw ambient light around the set the same way your finished replacement video would, a “3-d” effect that would cost you time and work to replicate in post, when you can get it for free on set as a “practical” effect. The TV is a practical light source on your set.

Again, it’s about a subtle yet key detail that “sells” the replacement shot. You may not, as a viewer, understand consciously why it works better, just that it does.

When you shoot the over-the shoulder from behind the TV reversals of God watching, you can use that same flicker source feeding the screen as a keylight. If God wears glasses, you’d better run the actual replacement footage full-screen on the TV set if you can.

[Mark Suszko]

If using existing motion footage, “posterize time” would do this.

For original stop-motion, or drawn-paper animation on twos, I suggest Boinx iStopmotion.

[Mark Suszko]

It’s not true 3-d, I like to call it: “2-and-a-half-D”. They are 2-d planes, like panes of painted glass, inside a 3-d volume of space, like a fish tank. But if you get the proportions and angles right, it fools the eye pretty well.

As to the fog/smoke, I usually buy some smokebombs every fourth of July and store them in a secure container in case I need smoke some other time of the year. To get the level of smoke and fog in “Get Low”, you’d probably have to use professional grade rented smoke machines, a brand-new, never-used, pesticide fogger, or military surplus smoke marker “grenades”. Not only that, but if you shoot that scene outdoors, the wind is going to rapidly dissipate the smoke unless the shooting space is confined with wind barricades of some sort. One of the reasons the scene is in slo-mo is that the smoke/fog only probably lasted thirty seconds or less in real time, before it blew away.

Ideally you would shoot that scene in a wind-less, indoor space like a hangar, parking garage, warehouse, something of that sort. And that much smoke or fog will certainly draw attention, maybe even false fire alarms, unless you take steps with the authorities first.

Things that seem simple, often are not.

[Mark Suszko]

Yes, you plug 1400 dollars into the account of a stage hand with a fog machine. ☺

You are not likely to be happy with trying to make this work in post production. But let’s say that for some reason, it just can’t be shot practically with smoke and fog machines on set. What would *I* try?

A lot depends on what your footage looks like. Did you shoot it against a green screen, or at lest against a white or black wall or curtain? Or do we first have to rotoscope the bike and rider out of their background? If you have not yet shot this scene, then get real smoke and/or fog for this and save a lot of grief.

I’m going to assume it was shot on a blacktop lot with a blown-out white light backdrop. Because that will make it less impossible.

First off, I would export my raw bike and rider footage from Final Cut into Apple Motion. You can also do this in AfterEffects, but I’m more comfortable in Motion.

The next thing we need will be some “plate shots”. That is, a background or two.

The plate shots here will be of smoke clouds, both still and motion footage. You can generate some of this in Motion using it’s particle system. Or find stock footage of thick, billowing smoke, with an alpha channel. You’ll need the alpha channel for this.

Now we set up the shot in virtual 3-d space, with the plate in the background, the bike and rider in the middle of the “sandwich”, and a front couple of layers of smoke.

We’re going to exploit a feature of the 3-d camera in Motion’s 3-d space. You can adjust the distance over which things are visible to the camera, kind of like depth of field/ depth of focus in a normal camera. Only in this camera, when an object goes out of the defined range, it doesn’t just get more blurry; it goes invisible.

So, if you set that distance right, then angle the plane with the bike shot and key-frame it to move in and out of the visible range, you can add keyed layers of cloud smoke to pass in front of and behind this to help make a smooth transition.

The biker footage would also benefit from key-framed effects regarding the application of blur and grain.

Without having actual footage to see and play with, this is about as far as I can go in describing my approach. It will take a lot of time and work to make this come off and it still won’t exactly look the same. If more than half of the terms I used sound like jibber-jabber to you, you don’t yet have the technical experience to make this work.

Now you see why renting the fogger was a better deal.

[Mark Suszko]

People with money solve this problem with a motion-controlled camera setup that can repeat the exact path and camera settings in realtime from take to take.

People with less money use planar tracking to generate the data that then moves the object in synch with the pre-shot background.

Or…make the object in CGI, apply the motion track data to that, build a composite that way.

If you could describe your intended shot in specific detail, I could give you a better answer.

[Mark Suszko]

If it needs a graphic to make you understand what it is, that’s bad writing. It is not a crime to put the actual job you do in the name.

[Mark Suszko]

Thanks, Andreas. Took me a minute to navigate to it, but it worked.

[Mark Suszko]

Well, you could mitigate that fear somewhat, by making a render with just a still graphic and the audio. Doesn’t change the timing. I use Youtube to generate my transcripts a lot, since it’s free. But nothing’s *truly* “free”, in this case, you still have to go back thru and add capitals, punctuation, and clean up mis-translated words and phrases. But if your audio is very clean and clear, and the speakers don’t have weird accents, YT does a pretty good job.

If my audio isn’t the best, I might render out a version for upload with extreme use of the graphic EQ to knock down extraneous noise and get the speech as clear as possible. You only need it to generate the transcript, after all. No point in uploading the finished edit when a low rez version with filtered audio will do it.

[Mark Suszko]

I regret to say I don’t have permission to show you samples. While they’re technically public domain, accessing them outside the current user base is something I’m not authorized personally to do. That’s true for a lot of my work, stuff that will never win an award, but it does the job for the audience.

You can tell your boss that I’ve been shooting these kind of lectures for just shy of twenty years, starting on umatic tape and one-inch master reels, thru Betacam and DVC Pro and onto P2 and hard drives. And have done it the way you do now, as well as live-switching it on location and the way I’ve outlined in this thread. Before I had hard drive recording available, I’d shoot the screen with a dedicated second camera, or iso the laptop by connecting a $6k scan converter and portable Betacam deck or DVD recorder between the VGA output and the projector, which was better than the camera. I’d rip the DVD-R’s with MPEG Streamclip and drop that into my edit, aligning by hand using the audio. STILL faster than manually putting in every slide and adjusting the timing to make it work.

If you can’t afford a live-switch setup, ( and these days almost anybody can) in my professional opinion after doing literally hundreds of these, the most time- (and therefore cost) -efficient manner is to live-capture the slide output on an iso recorder with audio, and multicam edit it with the live camera of the presenter.

I’ve done about five of these this year so far, with run times ranging from 30 minutes to 4 hours and doing it with the iso recording, I’ve gotten the turn-around time down to about 1.5 times whatever the original running time of the live presentation was. That means I turn around a complete edit of a simple 4-hour lecture, with synched slides, in *about* six hours. Five, if I skip lunch. But that makes me grumpy.

And that’s with either one or two cameras (I run a locked-off wide shot on a second camera sometimes which is good for cut-aways and breaking up the monotony. It doesn’t add to the edit time.)

The way you’re doing it now, you are editing slower than real-time. I’m going to guess, your time ratio is about 2.5 to 4 times real-time. Meaning, an hour takes you two and a half to 4 hours to cut. Best case, you’re marking in and outs on the fly while JKL-skimming the camera track at 2x speed, you stop the playback, note the duration, and apply that new duration to the requisite slide, drop it in, move on. But that doesn’t get you the animations, unless you play out the slide show manually while listening to the camera track and manually tripping the animations on cue, screen-capturing that as you do it, as many times as that takes to get right, then saving the capture, then importing it into the NLE, and lining it up. Am I right? Now your edit time ratio’s about 5 or 6x, right? I could read a book by the time you get that job done.

I do a 4-5 hour annual training lecture for a certain state agency every year. Three years ago, we did it the way your boss makes you do it now, and the edit took about a week, more or less. Since we went to iso recording the laptop outputs and multi-camming it in post, we got the same edit job down to a single day, or 7 hours. That’s money. And we increased the quality from SD-grade to HD grade. The miles-long spreadsheet displays are now actually readable on the screen. Even after crunching it down to mpeg 4 compressed web streams. And we get time-sensitive product to the user about four days sooner than originally.

This year I might live-switch the same same setup into a Tricaster, and cut another two hours or so out of post production time. Plus I will still have the isos as a back-up to fix any mistakes in the live switch, unlikely as that is.

Seriously, you are throwing time and thus money away by NOT doing it with an iso record. Very worst case, the iso will show you the timing if for some reason you have to re-edit and drop in the original slides from a file. Copy-paste this post to your boss. I’m from out of town, as all the best consultants are, and by virtue of that alone, he’ll take my word over yours, though we’re telling him the same thing.

So they screwed it up once before. That’s the past. It’s much easier to do today with an all-digital workflow, and you need to do a dry-run test to be sure you understand the equipment and settings. You’ll need a distribution amplifier to split off the laptop’s HDMI output, to feed the hard drive recorder and projector together. For your first try, you might be able to rent that, and the drive. I like KiPros but there are plenty of others out there at various price points. Ideally that recorder needs to be able to accept a mic input for synch purposes, but if everything records free-run, you can still get by even without audio in the hard drive recording, and it will STILL be a faster edit than your current method… You’ll want a portable monitor for confidence monitoring of the hard drive recording. Drop 100-200 on a consumer HD camcorder and use that as a back-up iso of just the projector screen as well, cheap insurance.

There’s another way… I have never tried it because I don’t trust the laptops I encounter. But… if you were to turn on screen recording on the laptop just before the show starts… when it’s done, you have a live-timed animated recording that just needs to be converted to drop into your NLE timeline. I can’t know if a particular laptop holds enough memory to run the ppt and record the thing for more than an hour, so I can’t recommend it except as an experiment or for shorter projects.

And with that, another method comes to mind. Powerpoint has a “rehearse timings” presentation mode, that’s not as memory-intesive as screen recording. If you can get the lecturer to present with this mode, then save at the end of the presentation, you can tell powerpoint to output a movie with the rehearsed timings, which in fact are the live show timings…. see where I’m going with this? A variation could be, you bring your OWN laptop or tablet with a copy of the presenter’s PPT and during the presentation, in real time, YOU manually time out the slides following what he does, in “rehearse timing” mode. Take that home, render out the timed movie, (WITH the animations) plug it into the NLE, line up the start times and be considered a wizard. Test this before you count on it, as Powerpoint keeps changing from version to version and I don’t keep up with all the changes.

I hope something in there helped you out – if nothing else, to know that someone else understands what you’re going thru. Been there, done that, got the iso recording. 😛

[Mark Suszko]

I handle this a different way, maybe a little more “brute force”…

I bring an iso KiPro recorder to the event and capture direct from the laptop’s output (use a DA) before it hits the projector, with audio grabbed from the house system or my camera, which concentrates on the speaker. Captures all the animations. Preserves the exact timing for every slide. Auto-align them in post in seconds via audio match, and then I create a split screen/crop so the speaker and the slides are up together, in synch the whole time.

Then I can JKL skim the presentation at about double-speed and look for places that could use full-frame shots. In spots where I want the slide up full, or the speaker up full, I slice the tracks or mark in and out on the fly, and “normal” the needed segment, taking out any positioning and scaling, etc. with one click.

I find for the really long and tedious lectures, this is the fastest way and saves a lot of steps while delivering a good look that the viewer can appreciate.