I n t r o d u c t i o n 

           I've commented to some digital video enthusiasts (mostly on DV Central''s DV-L list) about my experiments  trying to achieve a "film look" using Adobe After Effects, and several have inquired about the paths I've  traveled down and to what effect. While I have not made any end-all-be-all conclusions on this subject, I've  found some decent results with a variety of techniques, from the fairly simple to the exceedingly complex. 

           I'd like to share some of them with those who are interested; hopefully to gain some feedback from those  who have either tried their own techniques or might simply react to my approaches, and help me find even  more effective ones. 

           I'm thinking this exploration might lead to a tutorial article at some point, but I consider this "research in  progress" as is.

 

S u p e r   B r i e f   B a c k g r o u n d :  T h e   P r o b l e m 

           Video looks like video. Film clearly looks different, even film transferred to video. There is a perception  among movie watchers that telecined film looks "better," more professional, and that video typically looks  "cheesy." Whether this is a consequence of historical psychological conditioning or is some true hardwired  scientific fact I can't really say. But I "see" it myself. Bogart: pro. America's Funniest: cheesy. 

           The actual differences in the imagery is a consequence of different technologies and can get extremely  technical. Put simply, it involves the photographic nature of celluloid and its exposure to light, and the  wholly different response of an electronic array and its sampled sensitivity to light. The differences manifest  as widely disparate ranges of contrast and color, as well as a distinctly different nature between film "grain"  and video "noise." And then there's "interlacing," but we'll get to that... 

           It is possible to compensate for some of this disparity with filtering techniques--altering the contrast or  gamma of a video image, and possibly adding blurred noise to simulate grain. (It is even easier, though  more expensive, to purchase DigiEffects' Cinelook After Effects plug-in, which I endorse for this half of the  equation. This plug-in features contrast and grain settings meant to simulate the quality of actual film  stocks, like Kodak and Fuji 35mm stock. It can be an effective and easy tool for contrast, gamma, and  grain.) The quality of the original image and a careful attention to the lighting conditions of the video shoot  can impact the success of this process dramatically. These techniques alone can make a static video  image look much more like a single frame of film. 

           Unfortunately, these image qualities are not the only difference our eyes perceive. While they may address  the "look," they don't address the "feel" of film in motion. (Cinelook attempts to handle this with something  called "Time Slur," but so far I don't think this half of the equation works all that well. I'll offer some of my  own ideas for techniques below.) 

           Video is shot at (roughly) 60 fields per second, a field being made up of either odd or even lines from a  single image; when you put two fields together (one right after the other) you get a full video "frame." These  perceived frames produce a moving image at about 30 per second. This "interlacing" of the image is what  makes thin horizontal lines on television "buzz," and causes other minor headaches for video artists  everywhere. It plays at least some small part in making video look like video. 

           Film, on the other hand, records its images in complete frames from the get-go, at the rate of 24 per  second. When film is transfered to video, a process called "3:2 pulldown" (or "cine-expansion") stretches the  24 frames of film onto 30 frames of video by repeating fields here and there in even distribution (one extra  field for every four original fields, to be exact). While some of these images--frozen by the pause button  perhaps--may now appear to have "field interlacing," it is important to remember that they actually started  out as 24 discrete frames. 

           Unaltered, raw, video always has 60 fields of information. That is, 30 sets of field-interpolated frames per  second. Film has only 24 (complete) frames. It is this disparity that causes the "problem" at hand. 

           [Note on Canon's new "Movie Mode":

           Recently, the newest DV cameras--like the Canon XL-1--have begun offering an alternative to "interlacing."  They call it "progressive movie mode," though in this case the term "progressive" is a bit of a misnomer.  What this "movie mode" does offer is the same moment in time in each field of video. So a frozen video  frame appears solid--taken from the same moment in time--and not "interlaced" from two moments 1/60th of  a second apart. While this does tend to make "movie mode" look "more like film" or "less like video," it has  its own set of artifacts. For one thing, the sampling of video shutters tends to be crisp, sharper than the blur  on 24-frame film. And since movie mode is actually sampling from time only 1/30th a second, the temporal  resolution is actually less than that of interlaced video.

           Which is not to say "movie mode" is not a step in the right direction--it is in fact a lot closer to what  experiments I've been trying to achieve with After Effects or other post-processing software. But true  "progressive" video--being a full 60 frames per second--would be even better. HDTV specs will hopefully  take us there in the near future.]
 

B r a i n s t o r m i n g   T e c h n i q u e s 

           In beginning my own experiments, I was unsure how much of film's temporal "feel" on video might derive  from the 24 frames being "expanded" to 30 and how much was simply eliminating the field interlacing to  produce solid frames (even if this meant 30 frames instead of 24). 

           There are essentially two approaches I took in trying to simulate the "temporal" qualities of film: 

           (1) The simple approach: Eliminate (or hide) the field-interlacing as much as possible to produce 30 solid  frames per second. 

           (2) Try to do the full run: Change 60 fields to 24 frames and then cine-expand these frames to 30 fps (60  fields). 

           This is an interesting problem that seems almost like an infinite loop. Think about it--you're starting with  video and ending with video, no matter what. 60 fields per second. You can throw out every other field  (show only the odd ones, say, twice--as both odd and even) and get 30 "perceived" frames, but your  playback is still 60 fields. You can blend and compress, or blend and expand, but you're still using 60 fields  at the start and producing 60 fields at the end. 

           What you do in the middle is apparently what counts.
 

T h e   S i m p l e s t   T e c h n i q u e :   T h r o w   I t   A w a y 

           I had been told (and had seen on several Net lists) that simply throwing away every other field (say, all the  even ones) would give you "film look." I was skeptical, to say the least. After all, that's half your image  information being tossed to the wind. 

           But I must admit this technique works, with some limitations. The resulting footage definitely gains a bit of  that film "feel." It moves more like film, less like video. Unfortunately, near-horizontal edges--already a  potential problem in video--become very stair-steppy with this technique. This aliasing problem typically  doesn't plague faces and human figures...but walk a set of stairs and watch your image buzz and quake. 

           Fighting this artifact led me to other simple attempts to lessen the aliasing. 

           (A   S t r a y   I d e a)
           I would venture to say that 16:9 capable DV cameras might actually benefit from this technique, if the  filmmaker chose to letterbox their results. (In other words, not take true advantage of the intended use of  16:9 mode, that of widescreen playback on 16:9 capable TVs.) In this way you could reduce the vertical  resolution (not quite in half, as you might want to do, but from 480 pixels high to 360 pixels high at least)  and not take quite such an aliasing hit by throwing away half the vertical info. Your image might soften a bit  as a result of the scaling, but the reduction in size would actually help make up for losing half the fields,  and thus half the information.

 

O n e   S t e p   U p   F r o m   E a s y :   B l e n d e d   F i e l d s 

           One thing I tried to reduce the aliasing problem was to blend my odd and even fields from each frame and  then hold that frame over both odd and even fields. This kinda gives a perceived solid frame and lessens  the effect of aliasing when you simply throw away half your information. It works best on longer (i.e. 1/60,  1/100) shutter speeds (blurrier footage) since the two images combine more distinctly into one blurred  frame. 

           In After Effects the easiest way to accomplish this effect is to use two different versions of your source  footage, one set to Even and one to Odd (Interpret Footage), and then treat these images as if Field  Rendering is Off. This way full frames are built up with every other line "interpolated" from adjacent lines.  These "interpolated" images can then be combined (50% of each) in a separate comp, either by using Add  mode with each at half opacity, or by blending one halfway (50% opacity) over the other at full opacity.  This final blended comp can be output without Field Rendering to produce final footage. 

           My results with this rather simple technique were pretty decent. Not as dramatic a "feel" as the "throw-away"  trick, but still more film-like and less stair-steppy than the original video.

 

T h e   D i f f e r e n t   S o f t w a r e   T e c h n i q u e   ( A E / E I ) 

           Another technique evolved from a fluke combination of different software packages. ElectricImage, a 3D  program on the Mac, has two different techniques for producing "motion blur" in its imagery. One is  algorithmically done at rendering time; it works somewhat well but has some bizarre artifacts and aliasing  problems. The more expensive technique actually renders multiple images for each frame (with no blur)  and then blends them together in some unique combination. How many multiple frames get rendered is  selectable (the more the better the results), and the amount of blend between them seems to be some  gaussian curve around the center, though I can't say for sure what EI is doing. (As an example, if you chose  5 frames, the 1st frame might be 10% opaque, the 2nd 20%, the 3rd (middle) one 40%, the 4th 20%, and  the 5th one 10%--totalling 100%.) Since the software is fairly automated, it's hard to know what the  percentages are. (You simply type in "5" for the number of source images to be combined into a single  image.) 

           Anyway, my partner (Jance Allen) and I decided we would use After Effects to produce a slowed-down,  blurred (frame-blended) series of frames (say, at 500%), and then use ElectricImage blending algorithm to  blend them back down to the original time (5 images become 1). The difference in AE's blending on  expansion and EI's blending on compression might add up to produce a better (or at least different) final  image. 

           Indeed, the results were very positive. 

           The final image moved much more like film. It is subjective, and some source material produced much  better results than others. (Some, in fact, looked no different than the original.) But we'd clearly stumbled  onto something that worked.

 

T h e   F u l l - B l o w n   A f t e r   E f f e c t s   O v e r n i g h t   R e n d e r 

           I have since tried to duplicate much of the success of the AE/EI technique using only After Effects. Having  to render an intermediary image (especially one 5 times as big) before passing it to AE is improbable for  long sequences. (And of course, my hope is to treat DV features with this technique!) It is possible to do this  but it's a sneaky setup in AE because you have to be sure you don't simply undo the time-stretch from earlier  comps when you recompress. 

           I've played with variations on this theme and ended up with some that work, even if I'm not sure why. (!) One  that produced interesting results is shown below. 

           So...The first comp is just the original footage, at 100%. This makes it easier to replace your source footage  later, I think. 

           The source footage is stretched 200%, turning each field of the original into a discrete frame. AE interprets  the missing fields from the odd/even data. Each new "frame" is softer, but it is distinct. 

           Next we blend these frames, five at a time. Each "frame" (originally a field) is offset in time one frame so we  can do some kind of "ElectricImage-style" gaussian blending (in the image above each is at 20%, but I've  also played with the 10%,20%,40%,20%,10% curve as well). 

           When we time-compress this blended comp in the next comp (40%), we end up at 24 fps, with the  information in each field being a unique blend of the original. 

           Finally, cine-expand these out to 30 fps, and make this movie. You can simply use the Rendering Options to  add the 3:2 pulldown and maybe eliminate a comp layer.

 

T h e   C h a l l e n g e   ( T h e   R e q u e s t ) 

           I am hopeful that progress like Canon's XL-1 "movie mode," and especially progressive-scan high-definition  television, will soon make this tinkering pointless. On the truly positive side, the current wave of DV  camcorders puts an unprecedented level of quality into the hands of anxious low-budget filmmakers.  Unfortunately, the psychological nature of "film" and its characteristic "look" make serious movie-making  with these cameras somewhat less credible. While I believe a meaningful, well-directed DV film can be  made and taken seriously, digital treatments like synthetic "film look" could make the playing field a bit  more level. 

           If anyone finds fault with any of the techniques discussed here, or has further ideas on how to improve or  enhance these treatments, I would welcome a dialogue on the subject. Perhaps an exchange of AE project  files is in order, or a public domain library for film look tricks and techniques. I welcome your input. 

           Please email zane@rzanerutledge.com and let me know your thoughts.
 

M o r e   T o   C o m e  . . . . !