Behind the Scenes

Making a Vision: How We Recreated the Predator's Vision
It goes without saying, but I'll say it anyway; you can't have a Predator movie without Predator vision. It is a key part to the real Predator films (you know, those ones that Fox made), and also plays an important roll in the comic books and computer games. So, when we set out to do both Dark Jungle and Hunter's Match, we knew we'd have to figure out how to make Predator vision.

How'd Fox do it?
The word on the street is that, in the real Predator films, they used a real thermal camera coupled with a regular camera so that they could blend the two images together to assure the humans stood out from the jungle the way they were supposed to, and that they were still definable as human bodies. I'm not positive on the price of a thermal camera, but I'm going to guess than one thermal camera would cost more than all of our fan films put together and tripled. So, we had to figure out a way to fake it. That's where photoshop comes in. Following is a detailed process of how we created the Predator's thermal vision thru the trickery of Adobe Photoshop.

Isn't Photoshop for Photos, not Video?
Yep, Photoshop isn't meant to be used for video special effects, but due to the nature of this one and the enormous number of steps involved, there's no video filter or effect you can apply that'll make thermal vision. That's why we rely on Photoshop, and you should too.

Tools for the Job
Personally, we use nothing but DV video, and percieve VHS and it's devices as demonic beings sent to degrade the quality of film and video. Don't bother spending the money to get your old VHS camcorder to work with your computer; buy a DV camera and make sure you have a Firewire port on your computer, plug it in and get to work. If you're serious about getting video equiipment, this is the place to start. You're also going to need some video editing software. We use Final Cut Pro exclusively. iMovie is the free and simple alternative to FCP, but it's obviously very limited in it's capabilities compared to FCP.

The First Steps
A few things to keep in mind when filming your Predator visions is that they must be well lit (I'll explain why later) and filmed so that it appears to be thru a Predator's eyes. We hand held all of our Predator vision shots, and spent most of our time on a ladder to give the illusion that the Predator was up in a tree looking down at the soldiers below. You have to put the cameraman where you'd want your Predator to be. You want to keep your video clip fairly short, no more than 6 seconds, because each second is thirty images, so you can see how quickly your workload will double with every passing second.

Once you have your Predator vision clip in your video editing software, you need to export the video clip as an image sequence so that the frames of the video are actually broken down into images. We use QuickTime 6 Pro to export our video as an image sequence, and you can get to the QuickTime export options directly from your video editing software. Be sure to set the image quality to the highest possible; you don't want to degrade the quality of the video.

Let the Photoshopping Begin
Ok, so you've got a couple-hundred images in a folder now, numbered 001-howmany; now the fun begins. If you have Photoshop 7 or later, keep the File Browser set to your images folder so you can quickly see what frames you have done, which one you're on, and how many you have left. Open up the first image in the sequence. Use the De-iinterlace filter (Filters/Video) to eliminated any torn figures due to fast movement. Duplicate your layer so there are two of your image stacked atop itself. Name the lower layer "Background" and name the duplicate "Layer 1" if it isn't already named that. With the lasso tool, carefully and accurately trace everything in the image that is supposed to be "hot", which would include all of your humans. I exclude their guns, but that's just a personal preference.

Having trouble?
If you're having a tough time differentiating between the background and your people, that's why I mentioned that the shot must be well lit; it's for your own good. Now, if your shot is like mine and your people are wearing green camoflauge in a green jungle, you just have to go thru some of the frames to figure out what's part of the people and what's the background; this is the most difficult part of the entire process. If you can get this licked, it's all downhill from here.

A Little Background
Now that you have your people(s) highlighted, save your selection as "figures", because you'll need to reload it again and again. Smooth your selection by 5 pixels, more if your people are close to the camera, and save that selection as "smooth". Invert your selection so that everything but the people are selected. Invert the image. Use the Median filter on a factor of 5. Adjust the Levels (Image/Adjustments/Levels) so that the background is darker; I changed the levels to 0, .75, 255. Create a new Layer and name it "background color". With only the background selected still, fill the "Background Color" layer with Blue, RGB values R 0, G, 33, B 87. Set the "Background Color" layer to "Color Burn" from the Layers Pallette. Your background is now complete.

For the People
If your selection is still set to the background only, inverse it again so that the people are selected. If you no longer have your selection, load your "smooth" selection. Return to "Layer 1" and Invert the people, give them a Median filter on the factor of 3, and more if they are closer to the camera. Create a new Layer and name it "Highlights". Be sure your selection is still "smooth", and go to "Color Range" under the select menu. In the dialog box that appears, choose "highlights" from the "select" menu, and click "OK". This will cause your selection to change so that only the highlights of your people are selected. Fill that area of the "Highlights" layer with red, RGB value R 237, G 28, B 36. Set that layer to "Saturation" from the layers Pallette. Turn off the "Highlights" layer with the eyeball next to it.

Reload the "smooth" selection. Create a new layer and name it "midtones". Return to the "Color Range" box under the Select Menu and choose, you guessed it, Midtones. This will select the midtones of your people. Now, fill the "Midtones" with Green, RGB value R 57, G 181, B 74. Set this layer to "Overlay" from the layers pallette. Hide this layer for now.

Reload the "smooth" selection. Create a new layer and name it "shadows". Return to the "Color Range" box under the Select Menu and choose Shadows (surprise!). This will select the Shadows of your people. Now, fill the "Shadows" with Blue, RGB value R 0, G 33, B 87, the same blue you used for the background. Set this layer to "Color Burn" from the layers pallette. Lastly, create one more layer and name it "All". Reload the "smooth selection" and fill the selection with red, RGB value R 237, G 28, B 36, and set the layer to "Difference" from the layers pallette. Turn all of your layers back on to see what you have created!

The Fine Touches
With your "smooth" selection still set, return to "Layer 1" and adjust the Levels (Image/Adjustments/Levels) of the people to your personal preference. I changed the levels of this image to 0, .60, 255, but every scene will require different level adjustments. Lastly, you can blur the edges of your people and add a blue ora around them to take some of the sharp edge away. Click this image to see a larger view of the completed frame.

You're Insane
Congratulations. You have completed 1/30th of a second of video. Only 179 more to go to make 6 seonds! Ok, so you're probably wondering if there's an easier way to do this. Well, yes and no. First, budget allowing, you could buy duplicates of your characters costumes and paint one set blaze orange so that when you film the Predator visions, they can slip into their blaze orange suit, throw some orange paint on their face and hands, and they'll stand out in the woods like a nun in a porn shop. If the light is right, it will take but one click with the magic wand in Photoshop to seperate them from the background, and save you the tedious work of tracing around them manually with the lasso tool. Second, you can record every step after the selection as an "Action" in Photoshop, so once you outline the people, everything else will be automated. Lastly, because you saved the original selection as "figure" you can use that selection on the next frame, allowing you to simply alter the selection to fit the new frame; this also results in much smoother looking characters. To the right is an image of what your layers pallette should look like if you follow our steps.

If you know of a better way to do this, let us know! Until then, this is how SBOOB makes their Predator Vision. If you try these steps, we'd be curious to see your results. Teh Ned.