You would probably have to rig something with expressions - a linear wipe, that "crops" only 1 pixel plus some code on the position that offsets each layer by that 1 pixel column and even more expressions for time-remaopping so that each frame is correctly represented in its pixel column. That would be based on the layer index as the multiplier and as you duplicate the layer, it would offset automatically. It may just not be particulalrly practical to have 1920 layers to fill a HD comp or even more. And the expressions will make things slow to begin with. Other than that feel free to experiment with Time Displacment as already mentioned here: http://forums.adobe.com/thread/1068924?tstart=0
Thank you Mylenium, I was hoping it would be a more simple task, this is somehwat beyond my capabilities. Is there any other softwar eout there that would do this more easily?
You could also import the footage into Photoshop as files and do it manually. Actions might help speed up the process, but it might take a while to figure out how to get it to work the way you want it. Also, try asking on the After Effects expressions forum and see if anyone has an idea about how to set up an expression that corresponds to Myleniums suggestion.
I'm not aware of any software that specifically does this, but as Ben suggested, there may be a way with Photoshop actions as well. If you're not in a rush, I may give this a try tomorrow and come up with something in AE. It's too late in the evening now here and my brain is munched...
Thank you Ben, I had thought of doing it in Photoshop but was worried it would be impossibly time consuming. I will follow your suggestion and ask the effects expressions forum.
I'm not exactly sure what you're trying to accomplish here. Could you describe the shot or process a little better.
What I think you want to do is to take a single pixel vertical strip from frame 1 at column 1, then take another column of pixels from column 2 and frame 2 then do it again for column 3 frame 3, column 4, frame 4 and so on until you have one column sampled for each frame for the entire composition.
I think that you then want to stack up these columns of pixels to create a single frame with each column being a different slice of time.
You can do that with a gradient and Time Displacement. Just select a black to white gradient with black on the left and white on the right. Pre-compose that layer. Then take your footage and apply Time Displacement. Set the CTI indicator to the frame that is half the width of your composition. IOW, if the comp is 1920 X 1080, set the CTO to frame 960.
Now adjust the time resolution to your frame rate and set the offset to the position of the CTI (960) by the frame rate (30) and you've got it.
Frame 960 will have a column of pixels from each of the 1920 frames in your movie.
Here's a screenshot of a single pixel line moving from the top of the frame to the bottom of the frame with time displacement applied. The comp settings are 872 X 486 @ 30 FPS.
The gradient in the background is a pre-comp that is driving the Time Displacement.
That certainly seems like a much better way of automating that process than doing it manually in Photoshop.
I am not in a rush today Mylenium and would very much appreciate you having a shot at it in AE.
Thank you, Nicola
Did you take a look at the Time Displacement solution. It's every bit as accurate as duplicating a layer 1920 times (for an HD comp) and fussing with the time and the slice with expressions.
I still am not exactly sure what you are trying to do other than create a single frame with a bunch of time slices.
The Time Displacement solution that I offered would also create a movie with this multi time slice image moving at the same rate as the movie that generated it.
I have to agree with Rick: what visual look are you trying to accomplish? It would take A LOT of frames to create a single frame.
In my mind's eye, your process as described would yield a bizarre-looking, non-sensical single frame. The ultimate look would depend on the motion in the original footage, of course.
Or is this more of a, "Can AE do this?" sort of question? If so, you have a few techniques offered.
Thank you all for your input, here is more info about what I am trying to do:
I am a photographer and I am trying to create a finish line type image from video frames, so taking a single pixel vertical strip from the same place on every frame of about 1000 frames and then lining them up in order side by side to build up a realistic image.
I have a few challenges:
- the first in capturing; as my subjects are moving at approx 60 miles per hour and I will struggle to get the frame rate (and enough light through a long lens) required to get enough 'slices' to make up a realistic image.
- the second challenge is in the processing; while I am pretty familiar with Photoshop, I have never used AE and so before I commit I was hoping you guys would be able to let me know if this is the right software to use and if it is easy enough in AE for me to actually do it.
I don't think you'll get the kind of image you want by taking 1 pixel wide slices of time. You'll end up with a time smear.
Take a look at this short video of someone walking using the time slice method you're describing.
If you have a sample of what you want it might help us to point you in a more efficient direction.
...and again, I'm in agreement with Rick: I don't think you'll get what you're after.
You say you're shooting video, and you say the subjects move at about 60 mph.
At what frame rate do you intend to shoot? The fastest frame rate on typical video cameras is 59.94 frames/ sec.
Let's say you shot a still at 1/60 sec.
If you hold the camera still, the subject in the shot will have motion blur.
If you pan the camera to follow the subject, the background will have motion blur.
Does this constitute what you are calling a realistic image?
I'm still having trouble understanding how the technique you propose will yield anything useful. It certainly won't look realistic.
I am trying to get the step before the video example you have shown above. This chap has taken vertical pixel line 1 from each frame of a film and lined them up to make an image that then becomes frame 1 of his newly composed film, he then moves onto pixel line 2 of each original frame and lines them up in an image to become frame 2 of his new film etc. I am only trying to make the one still from the same single pixel line of a few thousand frames and line them up. It should looke something like this, http://www.artknowledgenews.com/2010_07_13_00_17_51_adam_magyar_to_photograph_hundreds_of_ people_walk_the_walk.html
Of course this involves people walking which is where the unrealistic movement shows, the bus in the background, as the bikes I want to shoot, remains relatively 'normal' looking. This is also shot at a much slower speed than I will need to use for bikes travelling at 30mph.
In the capturing it is going to be a balance between image speed, frame rate. I am happier with this side of it as I know a speed of 1/6000or 1/2000 of a second will capture a sharp image. The frame rate I am hoping to shoot at is 200fps and if possible 500fps for the few second burst I need and then extrapolate up to 1000fps or more with Twixtor.
The processing is really where I am looking for the help. I could open the film frame by frame and manually extract the pixel lines one by one, but with several thousand per image I am hoping a more automated solution is possible. I am not code friendly and so cannot write anything to do this that way, so was hoping AE might be able to help.
You are misinterpreting the technique that is used here. The photographer is using a slit scan panoramic camera but is panning at the same speed that the people in the shot are walking.
These cameras have been around as long as there has been photography. The film back is curved and the lens and a slit at the focal plane move by a gear mechanism moving the slit over the film. We have a photograph of my great grandfather in Park City, Utah in the 1880's shot with one of these cameras. It's a group shot of all the miners working at the mine but great grandpa is on the left side of the group and also on the right. He just ran behind the group after the lens passed him and got to the other side before the lens did. They used something like this:
Here's one from the 40's
Normally, the camera is stationary and the lens and slit moves. In the case of your example, the camera is panned in the opposite direction the lens is moving so that the moving subject, the people, stay centered in the frame, and the background is blurred. This also isn't a new technique. There's a very famous photograph of the 1948 olympics that was shot like this.
You can achieve this effect easily with a standard video camera and Time Warp by simply panning with the subject. It's dark outside now, but in an hour or so I'll shoot something with my iPhone and post an example. With a little clever pre-comping I can even simulate the panorama. You may even be able to do this with the new panoramic shot feature with the iPhone iOS6. I'll even give that a shot. You'll never get this effect with a high shutter speed, a stationary camera and thousands of frames.
In 1974 I did a story on a Seattle firefighter that built his own camera that used 10" wide roles of aerial photography film to shoot amazing panoramas of Seattle.
Here's a panoramic camera you can buy today for $275 that will do what you want to do.
Well, all that theoretical stuff aside, here's what I've cobbled together:
http://www.mylenium.de/stufffiles/look_color_strip.zip (CS4 and above)
Maybe it's of some use, after all. Though, as Rick and Dave said, if you really want to produce a slitscan effect, you'll have to do a lot more than just slice up a sequence of film frames.
I had a thought this morning about achieving the effect with a video camera doing a pan. In that case the frame rate is the problem. No matter what you do you're only going to have as many time slices available that the camera can produce. If you wanted a 1920 slices of time over a 3º pan then you would need to shoot a 1920 frames or about one minute and 4 seconds at standard video frame rates. If your camera was positioned so that you would see 50' of the finish line and your subject was traveling 60mph (88 feet per second) it would only take about .56 seconds to cover the field of view you are interested. That means 1920/.56 or about 3300 frames per second. At that frame rate you wouldn't get the nice blurred background, but you could get the same distortion you'd get with a slit scan panoramic camera panning to follow the action.
I did check out the iPhone slitScan camera option but it only samples about every 1/10 of a second so that doesn't work either.
Mylenium.. I checked out your project file. I think it's broken. The Color strip footage never moves into the frame and isn't offset as layers are duplicated.
Mylenium.. I checked out your project file. I think it's broken. The Color strip footage never moves into the frame and isn't offset as layers are duplicated.
Works fine for me - getting nice temporally offset strips from the colored balls and all sorts of funky patterns and loading in some footage in the source comp also makes it beautifully striped....
Ah, my bad. Zipped up a prototype version with a variable angle control that I didn't finish. Correct version now in place. Please download again. Sorry for the confusion.
Thanks for this sample. That's very clean work. Render times, as expected, are fairly imposing.
There are some really good ideas in your expressions.
I also just wanted to add my favorite panoramic camera, the Soviet Horizont. Here's a really good condition one on ebay at Buy it Now for $300.
A bit more info about the camera:
Mylenium, thank you. From your sreenshot, it looks like what I am hoping to acheive is in the tweaking of these settings; I just need to get my sleeves rolled up and get in there and try some things out.
The time to render is not a problem for me, but the Mac power and memory may be as I will need something crazy like 2000fps to get the image quality I am looking for. Do you think this is even do-able?
Thank you again for your help, Nicola
Thank you, Rick,
If I understand what you are suggesting correctly, I think I am going for a different and third option here; the camera stays still, the slit stays still, and only the bike(s) zip passed in less than a second. So if I can get a fast enough fps speed then ideally nothing pans or moves other than that bike. This should lead to the background is continuous stripes and only the bike is in sharp (hopefully) focus.
As youhave suggested, it is easy-ish to do with a film camera and the film being run behind the lens at a similar speed to the moving bike, but with the speed of the subject it produces serious vertical banding in time with the mechanical 'pull' on the film and the horizontal lines are pretty wavy as it is pulled so quickly and mechanically through the camera body.
I could also create a ''Photoshopped' solution by making a background and dropping a sharp bike on it (this would be by far the easiest solution)...
...but I am all about 'in camera' and digital and I know it is ambitious, for all sorts of reasons, but I am not quite ready to throw the towel in on an honest digital solution.
If I could see and understand a low res, low fps version of what I am hoping to achieve (and I think Mylenium's solution is on the right road for this) then I would happily plug away at the higher res and higher frame rate challenges.
Thank you again, Nicola
1 person found this helpful
Mylenium's solution is very clever but it's not going to give you any different effect than Time Displacement and a gradient.
If the bike is moving and the camera is still then the time slices will slice up the bike into strips and keep the background the same. To make a complete bike you need to be panning with the bike. When I get a moment I'll post a shot that shows what I mean. There's no way to hold the camera still, have the bike move and then slice up time and get a complete bike. Each time slice will only contain a small slice of the bike.
There are digital slit scan solutions out there. They work exactly the same way a slit scan film camera does. You just can't do this with technique with a standard full frame camera.
Here's the shot with horizontal time slices (because the cars are moving toward and down in the frame.
Notice how the car in the middle lane on the left is chopped into a bunch of frames but the cars in the right lane moving away are shortened way up.
If I get a chance later tonight I'll slice up a shot where the camera is moving. The subject will be relatively undistorted but the background will be cut up into slices. If you look closely at the image above you can see he time slices in the center barrier and the lane markings caused by camera movement.
I am thinking the key is to get the slices so small/short (1 pixel wide ideally) that they are difficult to see, so a ridiculously fast shutter speed and high fps. In the real world, if I get 200fps at 1/1600s at a less than super grainy ASA, I will be doing very well. With the bikes doing 60mph, by my maths, I am still going to be short of frames to render the bikes in realistic proportions. I will try Twixtor and see what the quality of infill is like, but...
I am coming round to agreeing that some amount of panning may be necessary and even beneficial. I just don't want to go for a full pan at the speed of the bike or there will be no reason for a slit composite / time displacement; a pure pan shot is nothing unusual or different. But maybe a balance with some pan; and so at slower fps the background and subject would both be 'sliced', but at higher fps speed could acheive the sharp bike and smooth striped background.
Shooting the the super slo-mo to get the 200fps will of course yield some fun and interesting footage in itself.
When you did your shot above Rick, were you using Mylenium's settings? He was using a 20 pixel strip width which will clearly show the fast moving car as strips, if you go to a 1 pixel strip, the distortion will be greater but does the image smoothness improve?
Thank you gain for the help and support guys,
With a 1 pixel strip you would still see 30 pixel tall slices because the video was 30fps. It would be meaningless to slice any smaller.
In regards to your wanting to get a sharp image of the bike I think you still are missing something. If the camera is not panning or moving with the bike at approximately the same speed then you're going to get slices of one part of the bike only.
For example, if the camera is still and it moves from left to right and it takes 1 second to cross the frame. Let's say your time slices see a strip of the bike that's 1 inch wide. The resulting image would be that one inch of the bike repeated all the way across the frame. It would be just like the image of the car above that's not really a car, but just a portion of the car repeated many times.
This would be nothing like the image you showed us as an example. There's no way of generating that kind of image without approximately matching the speed of your subject.
Shooting at a high frame rate without panning would only give you smaller strips so the portion of the bike in the strip would be narrower.
You'll end up with something like this:
The single column of pixels just right of my daughters nose is repeated across the frame. That's about what you'd get with a stationary camera.
This is a very weird hybrid manual/digital technque I was playing with a lot last year to make aniamtions: Try taking a sharp image of someone riding the bike with a very fast shutter. Then move it across the bed of a scanner While it's scanning so that the scanner captures and distorts slices of the image as it moves. Depending on how fast and when you move the image with the scanner you could potentially control how distorted the background is in relation to the subject. I know this is a totally different technique than what you were originally after, and it will no doubt render different results, but I feel like it might get you in the ballpark of what you want. Also, one thing I realized when doing this is the higher you set the dpi on the scanner, the slower it will scan. You could also shoot multiple pictures to create a panorama the size of Adam Magyar's work with a normal DSLR camera on a panning tripod head. Then assemble those images in Photoshop with Photomerge, print out your panorama and drag it across the scanner to distort it. Here's an image I made last year by just moving my face across the scanner bed :-)
The scanner only focuses on what's directly touching the bed, but if you had a photo you could hold it down is you move it so it would be sharper. As I said it's a totally different technique, but it's fun and might give you more control.
Not to beat a dead horse but this is what I got when I shot a short video of my daughter riding beside me. Sliced it up into a bunch of vertical slices in time. You can see the wave created by her knee, and the extended arm because we were not perfectly synced for the 6 seconds of this shot:
For even more fun take a peek at this video. You can see the slices develop over time. If I was perfectly tracking her head for example, it would be relatively distortion free throughout the video but the wavy image of her knee would remain.