Special note: The
information contained in this document is written to the best
of my knowlege. I take no responsibility for any mistakes
that occur as a result of following anything here. So don't
sue me if you blow a million dollar budget because I made
a typo. Remember the cardinal rule of filmmaking. (Well, maybe
not the cardinal rule, but an important one anyway): Test
everything. If you plan on doing a feature and are going to
edit it with a process you've never tried before, you better
shoot a test roll and take it through to the final format,
whether film, video, etc. There are so many places to screw
up that a test is an absolute necessity. I know people who
have ignored this and decided that they would "figure it out
as we go" and ended up out of sync or with unrepairable problems,
so listen to me. Please.
Over the past year or so, a number of
producers and directors have sent me email asking how to
edit a film project on video. That is, they want to shoot
film, edit the film on video or a computer, do sound design
in video or on a computer, and then have the final product
end up on film again.
Why not just edit on film?
Anyone who has edited with a Moviescope
and squawk box is familiar with the pitfalls of film editing--
it is a slow, confusing process. You have to be incredibly
careful with trims, frames, labeling, storing, etc. It takes
forever. It looks and sounds like crap as you are editing.
In short, the cheapest way to edit, by cutting and pasting
actual work print, sucks.
Flatbed editing isn't much better. Although
you are sure that the film is playing at 24 frames per second
(fps) you still have many of the same problems with workbench
editing. You lose trims, you have to label everything, you're
editing sound on magnetic stock, edge coding is a major
pain in the ass.
Now, don't get me wrong. Many die-hard
film editors say the only way to edit a film is on a flatbed.
They say that the slowness of the process forces you to
examine each shot precisely, that the hours of time spent
shuttling through rolls of work print results in an organic,
complete undersanding of your footage and will therefore
result in a better edit. As far as I'm concerned, this is
akin to saying that typing on an old-style typewriter makes
you a better writer than if you wrote on a word processor,
but hey, to each his or her own.
Technically, flatbed editing is non-linear
editing. Unlike in video editing, if you want to go back
and add a shot in a workprint, it's not too much of a problem.
This is a major disadvantage to video. The trade off is
time. A video editor putting together a video project can
generally move pretty fast on his or her own, as opposed
to the slower process of editing on a flatbed. So, while
flatbed editing is non-linear, it's not exactly what we
mean when we say "non-linear" editing.
Both flatbed and video editing share
some similarities in that the source material and final
edit are in a form that requires constant rewinding and
shuttling and stuff.
The Basic Idea
Ok, so what we want to do is transfer
the film negative to video, where it can be edited on video,
then somehow get the information off the video so that the
film negative can be conformed exactly like it appears on
the video. Finally, we're gonna have to have a way to get
the sound off the video and into a form where it can be used
with film.
The one other possibility we might want
to do is to output the picture, sound, or both to computer
where we can do the editing non-linearly (as opposed to
on video), and then return the information to film from
the computer.
Understanding the Film <-> Video
relationship
Let's start off with some basics relationships
between film and video.
| 16mm, 35mm Film: |
24 fps (frames per second) |
| NTSC Video: |
29.975 fps (usually rounded up to
30fps) |
The telecine process is the process whereby
film is transferred to video. There's a big machine that
costs a lot of money, and they throw the film on there,
and hit record on a video machine, start the film rolling,
and it records the film on tape. The machine commonly used
for this process is called a Rank Cintel, and so often the
telecine process is called "ranking" the film.
Ok, now since we'll want to edit on video
and then have the video refer back to the film, so we need
to be absolutely sure of which frames we're editing in video,
and how they relate back to the film frames. Video is about
30 fps, film is 24. You could run the film at 30 frames
per second while you are recording onto tape, and that way
you'd have a 1:1 relationship between film and video. But
of course, when you watched the video tape it would be running
too fast to edit.
There must be a way...
Obviously, you can't record one frame
of film to one frame of video, cuz the fps difference will
make the film go by too fast if you watched it on video. So
you gotta do a conversion.
In order to make a video dub of the film,
and stay in sync, the telecine process copies a frame of
film onto two FIELDS of video, then the next frame is copied
onto THREE fields of video, then two, then three, again
and again. This "pulldown" process assures that when you
watch the video version of the film, it's in real time (or
very close, see below)
Got that? It's important, cuz if you
don't have this exact relation between the film and video,
as is done in some transfer houses where they just run the
film and hit RECORD on the VTR, then when you match the
sound back to film you'll have major sync problems: The
Rank might not be playing at exactly 24 fps, it might be
slightly slower or faster, which will result in sound drifting
out of sync over an extended period of time. Or it might
vary speed, which is even worse-- the point is that this
rank must be very precise. You are after all editing your
film's SOUND and PICTURE to this, and it better accurately
represent the timing of the original film or when you go
back you're in trouble.
Pulldown: 30 FPS or 29.975 FPS?
Ok, now let me throw one more wrench
into this whole film-> video conversion-- Don't tell
anyone, but video isn't REALLY 30 fps, it's really 29.975
fps. So when you're doing this 2-3 pulldown in the rank,
you're slowing down the actual program length (cuz when
you play the tape back, it's playing ever so slightly slower
than the speed that the pulldown was calculated for).
(If you're familiar with drop frame vs
none drop frame timecode, you'll be familiar with the drop
frame -- a frame is dropped every few minutes to compensate
for the 30/29.975 fps difference in order to keep the TC
accurate to real time)
This means that when you edit the sound
on the computer or on video, you're actually editing a show
that is about 1% slower than the original film. If you were
to just lay that sound onto an optical, you'd be in trouble
as the sound would slloowly drift out of sync in the film,
even though it looked great on video.
When you put your field sound on the
ranked video tape, you'll need to make this compensation.
You can do this in a few ways. Protools or Deck II (and
I imagine Sonic Solutions) will do it for you (there is
a "pulldown" option) Many DATs will do it (change setting
from 44100 samples per sec to like 44110 or something (can't
remember the exact # right now)) or if you're used to a
Nagra, use an external crystal for this purpose. Oh, and
there's a switch on the Avid's SMPTE slave driver (Digidesign
hardware) that switches from .99 to 1.0 speed for this purpose.
Ok, let's review this: Because the ranked
video is not EXACTLY real-time compared with the film, when
you add your sound to the video transfer, you're going to
have to slow it down slightly.
For more information on 2-3 Pulldown,
see the detailed discussion of 2-3
pulldown.
Isn't there an easier way?
Well, as you can see, the whole relationship
between film and video is kind of confusing, especially
when you take into consideration the problems that occur
with keeping the sound in sync.
I've gotten a few letters from people
who have asked me if it's possible to just digitize the
film to the video and then do an edit of the video and send
that off to the conformer, and have them use the video timecode
to edit the negative. They tell me that there is a special
formula that you can use to convert raw timcode numbers
into film frames without having to do any of the pulldown.
The formula they propose usually goes something like, "As
long as you're cutting on frames ending in 00 or 05, it
will work. "
A problem to consider is that when you
edit frames of video, you are actually editing slices of
time which are thinner than the frames you started
with-- that is, one second of video is sliced thirty times.
Film, only twenty four. So when you try to match back your
edits to film, you end up with that problem of possibly
having included a fraction of a film frame into your movie.
The "formulas" of cutting only on certain
frames of video just don't seem to work. If someone can
relate a story of such a formula being used successfully
in a feature film, I'd
love to hear about it.
Similarly, using a burn-in on the video
screen (which indicates the original frames of film) and
editing with that can give a negative conformer an IDEA
to help put together the movie, but it will still be possible
to have partial frames of the film appearing in the cut.
Or, to look at the problem another way-- remember that when
film is transferred to video, two out of five video frames
will contain a single field from one film frame and a single
field from another. It's a video frame with a split per
sonality-- half is from one film frame, half from another
one. So what if this frame is included at the end of an
edit? It's confusing for the conformer, and certainly not
a precise way of editing.
So what to do?
Ideally the trick is to get the pulldown
to work correctly so that you can edit the film at 24 frames
per second.--- so that the frames that you are editing on
the computer have a 1:1 correspondence to the actual frames
of film on the negative. This is accomplished easiest on the
Avid Film Cutter and Media Composer with Film options, although
it can be done on other non-linear editors as well. I'm told,
for example, that Premiere 4.x can edit at 24 fps.
I have received email from other filmakers
who are trying different ways to get the picture into this
24 fps format. One user tells me about a program that can
supposedly convert 30 frame Quicktime files into 24 for
editing in Premiere (a relatively inexpensive nonlinear
editor):
>Ron,
>
>I just discovered that the latest version of Movie Tools
can do the reverse
>pulldown conversion (from 30 fps back to 24)! Unfortunately,
it's $195,
>but if that's what it takes, I'll do it.
>
>Movie Tools is a QuickTime movie processor, and in addition
to the
>30/29.97--conversions it can interlace/deinterlace,
performs alpha channel
>stuff, some different effects...a fair amount of different
processing
>functions.
>
>The address for the main site is:
>http://www.McQPro.com/index.html
Can't I just edit the damn thing on video
and let the conformer fix it?
It should be pointed out that some conformers
are willing to take a video transfer of a film that has
been edited on video and use it as a guide to conform the
negative. When they get to a problem having to do with a
fraction of a film frame being used as a full frame or another
ambiguity, they just take a guess, or use their own judgement
and add or subtract a frame to keep everything together.
If you choose to edit a project like this, it's important
to make sure that you talk to your conformer and decide
together how you want to handle situations where the cuts
don't line up exactly. It's also important to pay attention
to how your film got on video, so that the sound will sync
correctly when you put your video-based soundtrack on the
conformed picture.
I should have said it before, but didn't:
The "guessing" process of taking an edited video and conforming
a negative to it is called matchback (Hopefully you
haven't started conforming your negative without knowing
this word. Thta's why it's important to read these documents
all the way through before starting anything...)
Incidentally, this is what was done with
the film El Mariachi which had been shot on film
and then later the negative was conformed to match the video.
For that film, the editor had on occasion re-used the same
video material more than once in his film. Though the convenience
of using shots multiple time allowed him to get more out
of his film, it posed a problem when doing the conforming,
as the footage only existed one time on the negative. The
solution was the (expensive) process of printing multiple
copies of the same piece of the negative. The sound, too,
presented a problem. Luckily (or maybe unluckily) most of
El Mariachi had been shot without sync-sound, so
it was kind of a moot point, and the whole thing had to
be redone anyway. The conforming and sound design for this
film turned a $7000 film into one costing over a million
dollars. So be careful! (More details the making of El
Mariachi can be found in director Robert Rodregez's
book, Rebel Without A Crew)
Other Issues
Talking about El Mariachi makes
me think of a few more issues which you must consider if
you're going to be editing your film on video or computer
and want to go back to film. Remember-- the rules of film
and video are different.
- In video, and on the computer, you
can cut a single shot into your project many different
times. On Film, you can use each peice of negative only
ONCE-- You must keep track of when you are "duping" your
film frames. Using the same footage more than once will
become an expensive disaster when you try to conform your
video back to film and realize "oh crap, I've already
used that peice of negative."
- Always remember that in 16mm editing
you need to make a "handle" of an extra two frames at
the end of each edit. When you conform the negative into
A & B rolls (16mm film is usually conformed into TWO
negative rolls) the actual hot splicing process "eats
up" a frame at each end of your cuts. So it's important
to remember that if you're using a shot, then cutting
away, and cutting back, that you don't cut right back
to the frame that's on the border, because that frame
will be used as part of the hot splice. This is an important
consideration, because if you forget about this, you may
find yourself off by a frame or two...
- When editing on video, you may be
tempted to add dazzling video effects like page turns,
DVE digital effects such as picture-in-picture, split
screens, slow motion, etc. Remember that most video effects
are not available (cheaply) in film. So, doing slow motion
effects require generating a new step-printed negative.
Watch as your costs skyrocket.
- If you want to do an effect such as
a dissolve or fade to black, which can be done in film
through an optical effect, remember to consult your conformer
and/or lab to determine if the effects have to be of a
particular length. For example, the printer used to create
your final print may be limited to dissolves of only 24
or 48 frames.
- Depending on how you go about editing
the film, it may be that any cut list generated would
have to come from reading the burn in off the screen,
which kinda sucks, specially if you're editing on a highly
compressed, low resolution version of the film on computer.
- So another consideration about highly
compressed, low resolution editing, incidentally, is that
often you end up using shots that are (A) soft, or (B)
include things like boom shadows and other things that
are hidden in digital artifacts or outside the "tv safe"
area of your video transfer, so you have to be REALLY
careful to pre-screen the source video and meticulously
eliminate bad takes...
- Finally, this should be common sense
to most editors, but remember to always include a "Sync
Pop" in your soundtrack that machines a particular frame
on your picture, so that when you match your soundtrack
back to your picture, you'll have a reference for syncing
them up.
These issues may be intimidating, but
if you keep your special effects to a minimum and seek the
guidence and counceling of your lab and conformer, they
are not insurmountable.
In conclusion...
Obviously this stuff can get confusing.
If you have a suggestion for a solution to these issues
that hasn't been covered, a correction to something which
is either misleading, incorrect, or missing from this document,
please let me know.
I'd also like to hear from any
editors who have successfully (or unsuccesfully) edited
their film on video or computer so we can all learn from
your experiences.
Good Luck, hope this is helpful!
Ron
Friedman
film & t.v.
print & journalism
music & audio
advertising
mediaibc