DV.com http://www.dv.com/PrintableArticle.aspx?

id=16690

Print this document

How to Make Video Look like Film

The Rev. John Jackman

Dec 28, 2002
URL:http://www.dv.com/article/16690
A lot of software products claim they can make your video look like film. Some video filmmakers swear by
such plug-ins, while others sneer, "If you want it to look like film, then shoot film." Getting video to look like
film is a complex and nuanced subject (see "Matching Digital to Film," Oct. '02 DV, for a discussion about
blending video and computer elements with film). It's also the kind of topic that lends itself to Internet forum
flame wars. Oversimplifications and broad generalizations abound. So where do we begin?

We need to differentiate the separate topics of transferring video to film stock, and making video look like
film when played on a television or a digital projector. A video-to-film blowup is a peculiar and expensive
pursuit. A discussion about the film look usually refers to creating video that resembles a fine film-to-video
telecine transfer. It doesn't usually refer to trying to make video look like grainy, scratchy film run through a
poorly maintained projector-although, of course, one can do that. By the way, I am using the term film look
in a general sense. FilmLook is a registered, trademark process of FilmLook, Inc. (www.filmlook.com), of
Burbank, CA. FilmLook is a realtime proprietary process available only from this service bureau. You've
probably seen a number of FilmLook'd shows with-out being aware that they were shot on video. FilmLook's
success with programs on HBO and other networks that formerly demanded film shows the process
works-and making your video look like film is both possible and practical. Let's first take a look at the basic
differences between video and film, which break down into temporal resolution (frame rate), frame resolution,
and color representation. Temporal resolution refers to the fineness of motion over time. Film is shot and
projected at 24 fps, the minimum rate that early filmmakers found best preserved an illusion of motion. The
basic shutter speed is 1/48 second, or 180-degree shutter angle. Video is shot at 30 interlaced frames per
second, which really means 60 fields per second, with a basic shutter speed of 1/60 second. Although video
represents motion much more accurately than film, we primarily associate video with news and sports, and
the less-accurate motion of film with drama. Frame resolution represents the fine detail each individual frame
can capture, and 35 mm film captures at much higher resolution than video. But although the resolution issue
is a major one for blowups to film (see the "DV-to-Film Transfer"), it's hardly an issue at all for a film-like
look on video. Once a 35 mm film is telecined, its resolution is 720 x 486 pixels. Color representation refers
to the way each medium captures a particular color in the real world and reproduces it on a screen. There
are real differences and a lot of variation between video and different film stocks. Many film stocks have a
gamma curve that makes dark colors darker than reality (and more saturated), and makes highlights brighter
and more pastel. Video cameras typically have a flat gamma curve and actually do a more accurate job at
representing a real-world grayscale (see Figure 1). Film's latitude-the contrast range that can be captured-is
still greater than video, but video cameras have been catching up rapidly.

What is involved with creating a film-like look? To get the best results, start long before you use any plug-in
or proprietary process. The key building block is to shoot video with film production values: fine lighting,
careful exposure, good composition, lens filters, and controlled camera motion. You must take as much care
shooting the original video as does a film crew on a $50 million feature. If you set your camera on auto and
wave it around like Uncle Harry at a family reunion, you won't be able to make your video look like film with
even the most sophisticated post process. In fact, this type of home video footage will usually look worse
than the original when you apply a typical film-like post process to it. Film-style production values are often
enough to convince the average eye that the footage was shot on film. More than once, I've been asked by
video engineers if a piece of my footage was shot on film because it looked so different from the flat-lighted

1 of 4 04/16/2011 09:51 AM
DV.com http://www.dv.com/PrintableArticle.aspx?id=16690

video they were used to dealing with. I can't overemphasize the importance of production values. Although
production values are a major part of the battle, they aren't the final obstacle. The video footage must be
deinterlaced and color-corrected to match the particular look that you're trying to create. Sometimes a subtle
grain also may be applied to the video source.

Deinterlacing is at the core of all post processes used to make video look like film. To capture a true
telecine look, the 30 fps interlaced (30i) footage must somehow be converted to 24 fps noninterlaced
progressive scan footage (24p), then converted back to 30i by the same process used in an actual film
transfer, known as a 3:2 pulldown. Printing three full progressive frames followed by two interlaced frames is
a long-established method for converting 24p to 30i (see Figure 2). However, this is a relatively difficult
mathematical process that is likely to aggravate motion judder. Another basic method for deinterlacing is
converting 30i into 30p. This is a much simpler operation that doesn't create motion judder, and many
viewers find it visually indistinguishable from 24p. However, the deinterlacing should be done without losing
resolution. Some software deinterlaces by simply duplicating one of the fields, effectively halving the vertical
resolution. Sony consumer camcorders also do this when they are operated at a shutter speed of 1/30
second. However, the Canon XL1, XL1S, and GL2 can run in a true 30p "frame mode," which has made the
cameras attractive to filmmakers. The best software deinterlaces by blending fields, which preserves vertical
resolution and areas with no motion. Many methods of blending fields are available, ranging from a simple
50-50 blend to technologies that analyze motion and only blend the areas of the fields where there is motion.
This last approach gives the best visual results (see Figure 3).

Once the film is deinterlaced, the last essential touch is changing the gamma curve. The effect of a subtle
gamma curve change has a huge impact in perception. Highlights become more pastel, shadows become
denser, and midrange colors appear more saturated. The colorist can then add a color shift to make the
image seem warmer (the Kodak look) or cooler. Saturation may be increased or decreased slightly to achieve
a particular look (see Figure 4). And that's it- you're done. What? No grain? Well, all right, you can add a
little grain. But make it subtle. If you watch the re-release DVD of E.T. The Extra-Terrestrial, you won't see
much grain. Film manufacturers have been trying to make grain smaller and less visible. Of course, if you
want to match the look of an older film stock, you'll need to replicate the type of grain it had-but grain
matching itself is a subject that needs a whole article. From here, you can go anywhere you want: bleach
bypass, the green cast of some Fuji stocks, you name it. If you want to create the impression of old
projected film, you can add the heavy grain, scratches, and dust that many plug-ins provide. You can add
gate weave. But avoid these additions if you only want to capture the look of a fine telecine transfer. As DP
Rick Clabaugh said to me when we compared the output of several plug-ins, "We want to capture the best
of film, not reproduce its problems."

I tested a number of popular software plug-ins and standalone products, as well as several homegrown
procedures in After Effects. Almost every product had some difficulty with the dreadful 30i-to-24p conversion
math, creating aggravated motion judder that was distracting to even untrained eyes. Most did a much better
job with simple deinterlacing of 30i to 30p, and that may be good enough. The majority of my test viewers
preferred the 30p results, which looked more like a good telecine, over the 24p results with 3:2 pulldown,
which had the jerky motion of the rate conversion. All products offered effective but different tools for color
alteration. If you must use 24p with 3:2 pulldown while originating in NTSC, be aware that the camera
motion rules for film apply twice. You'll need a copy of the ASC Film Manual (ASC, 2001) by your side
whenever you shoot. You'll be more successful using 24p with 3:2 pulldown if you start with PAL footage
rather than NTSC footage. One product to investigate is DVFilm Atlantis (www.dvfilm.com), software that
converts PAL 25i to NTSC 24p with 3:2 pulldown. For a more detailed overview of the plug-ins I tested, read
the "Film-Look Plug-Ins" sidebar on page 32. For every product mentioned, I suggest you download a demo
copy of a product and test it before you buy.

With a sharp eye and subtle treatment, you can make your video look so much like film footage that few
viewers will be able to tell the difference. You'll get best results from carefully shot footage that's well lighted
and exposed-and especially not overexposed. In my tests, hot areas of video ended up with a very
un-film-like glare and aggravated video artifacts. Some of the most pleasing and convincing color conversions
in my tests originated from footage with a slightly compressed exposure range. The most satisfying post

2 of 4 04/16/2011 09:51 AM
DV.com http://www.dv.com/PrintableArticle.aspx?id=16690

results generally came from smart deinterlacing procedures that created 30p footage with motion blur, but
retained the detail of areas without motion. The look of film created by 30p didn't have objectionable motion
judder. Whether you choose a proprietary service bureau process, a commercial plug-in, or a homegrown
procedure, you can make your video look like it was shot on film. However, doing it effectively requires a
practiced eye, a subtle touch, and solid craft.

Many film labs now offer a video-to-film transfer service, traditionally known as a blowup. A few years ago,
video blown up to film looked like, well, video. Raster lines were visible and jagged edges were noticeable. It
wasn't pretty. The process has improved dramatically over the last few years. Although most DV readers
have seen films shot on HD and transferred to film, not everyone has seen a recent, high-quality DV-to-film
blowup. There are basically three processes for printing a digital frame to film: the old-style kinescope, the
film recorder, and the electron beam recorder (EBR). The kinescope is basically a direct film of a video CRT
with no effort made to change frame rate or improve apparent resolution. This old technology produces
predictable, video-like results. The major advances have occurred in the use of film recorders and EBRs.
Some film recorders are essentially very high resolution CRT displays the size of a 35 mm frame. The film
frame is exposed in contact with the CRT. The EBR projects an electron beam on the frame of film,
exposing it directly. The real magic, however, is what the software does to the frames before printing to film.
DV frames typically are deinterlaced and interpolated up to 2 K, or 2048 x 1556-pixel film resolution, creating
new data. All of the shops I contacted said they used 2 K rather than 4 K (4096 x 3112 pixels). Even
Kodak's Cinesite (www.cinesite.com) now does all digital film printing in 2 K, because the company found
that the 4 K resolution doesn't make a visual improvement in the output for most standard applications. I
screened 35 mm show reels from two DV-to-film specialty labs, Swiss Effects (www.swisseffects.ch) and
DVFilm (www.dvfilm.com). Both companies use film recorders at 2 K resolution, and their reels show an
assortment of scenes shot with a wide variety of cameras. Four experienced film DPs watched the
screenings with me. Overall, the results were impressive compared with the blowups I had seen a few years
ago. The frame-rate conversions from NTSC showed little motion judder. The spatial interpolation was good
and created clean lines rather than jaggies in most pictures. The old rule about "garbage in, garbage out" is
still in effect. Overexposed areas, particularly areas of clipping, displayed jaggies and artifacting. Ringing
(caused by a sudden transition from an overexposed area to a dark area) gave away the video origins.
Low-light and poorly lighted video translated into rotten-looking film. The quality of the camera made a lot of
difference, but the most obvious difference was the craft of the camera operator. PAL footage printed a little
sharper, and the best images came from a well-operated Sony DSR-500WSP PAL DVCAM camera. Although
the best examples looked as good as some film distribution prints, there were three main issues the film
DPs noticed. One was that the density of shadows was consistently poor. Blacks were dark gray, a little like
double setup on video. This is something that should be correctable through calibration of the film recorder
and application of a film-like gamma curve to the frame before printing. Another issue was that intimate
scenes worked better than panoramic shots, where the lower detail level was noticeable. Finally, skin
blemishes were more noticeable (as they usually are on video) rather than appearing a little airbrushed as
they are by the constantly varying grain structure of film. But, ultimately, these were minor if not subtle
issues, probably far beyond the eye or concern of an average movie viewer. Film blowups are expensive,
prohibitively so for the low-budget filmmaker. DVFilm quoted about $25,000 for a 90-minute feature; Swiss
Effects quoted $35,000. Labs that use an EBR such as the Arrilaser are more expensive, so figure $45,000
and up for the same feature. Nonetheless, these processes make it possible for DV productions to find
theatrical releases-at least until digital projection becomes common.

As part of the background research for this article, I tested a number of software packages, plug-ins, and
some basic procedures in Adobe After Effects and Premiere. Although this isn't a full-blown DV Roundup, I
will provide some general comments about the results. I used the same set of clips in each package: a
period shot of a woman reading to children by a fire, a fairly standard head-and- shoulders interview shot,
and a slow pan of a horse and carriage pulling up and stopping at a building. This last shot was the acid
test for motion artifacts because it contained a number of sharp, high-contrast edges on the carriage and the
building, the rotation of the spoke wheels, and the motion of the horse's hooves. The first two clips looked
pretty nice in all of my tests, but the carriage footage clearly showed any problems with frame rate
conversion and less-than-clean deinterlacing. Magic Bullet There is a big buzz about the Magic Bullet plug-in
from The Orphanage (www.theorphanage.com). You can read a full review of Magic Bullet in the last issue

3 of 4 04/16/2011 09:51 AM
DV.com http://www.dv.com/PrintableArticle.aspx?id=16690

(Nov. '02 DV). Magic Bullet ($995 for SD, $1995 for HD) is designed from a film DP's point of view-the
process and language is all film. Magic Bullet is divided into several modules, the major ones being the
deinterlacer and the color-corrector. The deinterlacer can convert either PAL or NTSC footage, and render
either 30p or 24p with 3:2 pulldown. The 24p/pulldown conversion was cleaner than in some other apps, but I
saw noticeable motion judder in the carriage footage, much more than would be in film. The 30p footage was
clean. The color-corrector, part of Magic Bullet's Look Suite, has a simple surface interface with a lot of
tweaks available under the surface. However, the Look Suite adds a heavy hit to rendering time. Magic Bullet
works in Mac OS 9 and OS X. A Windows version is in the works. CineLook and CineMotion CineLook
and CineMotion from DigiEffects (www.digieffects.com) are After Effects plug-in suites. Cinelook ($695
Broadcast, $1995 FilmRes) does color and gamma corrections, and also adds grain, scratches, and other film
flaws (see the review in Feb. '98 DV). CineMotion ($295) deinterlaces, adds 3:2 pulldown, and adds gate
weave and other motion effects. CineMotion also includes options for reducing grain, aliasing, and banding.
The two plug-ins together form an engineering playground, providing very sophisticated tools for grain
matching and reduction, as well as complex noise and dirt generation tools. CineMotion is a great product for
simulating or matching old film, but it has so many tools (some obscurely designed) that it may prove
baffling for many users and some of its power may remain unused. CineLook comes with a lot of presets
that DigiEffects designed to simulate particular film stocks. The presets are excessive, to say the least, and
produce looks that I can only describe as film at its worst. The DPs who also viewed some of these looks
had derisive reactions. I discussed my concern with Chris Athanas of DigiEffects a couple of years ago, and
he said that the over-the-top presets were a marketing necessity: "Customers with less experienced eyes
want to see a big difference." Fair enough. My suggestion to the more sophisticated user is to ignore the
presets in either package and start from scratch. Nudge the look gently in the direction you want it to go, be
careful with the grain, and forget the dirt and scratches unless you want them for a specific reason.
CineLook uses complex algorithms to generate extremely accurate grain, dirt, and so on. The program is a
legendary render hog because of this overkill. In fact, the rendering times of CineLook have been a strong
selling point for the realtime processing of the FilmLook service bureau because projects with tight schedules
can't wait for massive render times. CineLook and CineMotion work in Mac OS 9 and OS X, and Windows.
FilmFX FilmFX ($499) is a Premiere and After Effects plug-in. BigFX (www.bigfx.com) didn't re-spond to my
requests for test software, but I was able to test an older version at my "Film Look" workshop during DV
Expo West 2000. FilmFX was somewhat simpler than CineLook and CineMotion, but it had much faster
rendering times. FilmFX is currently available for Mac OS 9 and Windows. Twixtor The most accurate (yet
time-consuming) conversion from 30i to 24p was created by the Twixtor ($329.99 Regular, $495 Pro) plug-in
from Re:Vision Effects (www.revisionfx.com). This unusual time-stretching plug-in interpolates the motion of
frames and creates 24p footage with clean motion virtually identical to film's (see the review in Apr. '01 DV)
Re:Vision also offers the FieldsKit ($90), which deinterlaces to 30p, and ReelSmart Motion Blur ($90/$135), an
intriguing product for selectively introducing motion blur. ReelSmart Motion Blur allows effective simulation of
longer shutter speeds without affecting areas of the picture that don't move. Both must be seen to be
believed. These plug-ins are available for many compositing systems and editors running on Mac OS 9 and
OS X, Windows, and UNIX. DVFilm Maker DVFilm (www.dvfilm.com), a lab that offers DV-to-film transfers,
also sells DVFilm Maker ($95), a software package that adds warming and uses smart deinterlacing to
preserve detail in areas of the picture that don't move. The deinterlacing is clean and the effect is pleasing;
but curiously, there is no gamma curve control and the end result still has basic video colorimetry. DVFilm
Maker is available for Mac OS 9 and OS X, and Windows. Homegrown Workarounds If you use
deinterlacing in an NLE or appli-cation such as After Effects, you are throwing away half an image's vertical
resolution, which isn't a satisfactory solution. The After Effects Reduce Interlace Flicker filter does a fairly
good job if you check Motion Detection. One homegrown technique that works fairly well is to duplicate the
footage, place a lower-field-first version on top of an upper-field-first version, and set the opacity of the top
layer to 50 percent. Some vertical detail is lost, but the results are better than simply removing a field. The
After Effects Curves effect achieves most of what any of the plug-ins do for color, and the After Effects
Film Grain effect is useful, although it doesn't offer as many options as CineLook.

Print this document

4 of 4 04/16/2011 09:51 AM