Non-judgmental Preservation (was: chicago film archive)

From: Jeff Kreines (email suppressed)
Date: Wed Jan 03 2007 - 15:01:39 PST

>> For some reason hearing about grant money always makes me think of
>> how it might otherwise have been used, or more exactly, about what
>> factors go into choosing to preserve old films and which ones. I
>> wouldn’t conclude it should be used any other way, but what
>> exactly goes into this embalming process?

A good question. So many films, so little time, and never enough

Right now, it's very expensive to preserve a film, either
photochemically or digitally, and even large, well-funded government
archives might only preserve 50 films in a year. (And, of course,
restoration -- fixing problems -- adds greatly to the time and expense.)

This means that most of what is preserved falls into three categories:

   * Films that generate income, and are preserved by studios as an

   * Films that are "important" -- artistically or historically, and
have slowly risen to the top of a to-do list.

   * Films that are pet projects of curators (often for excellent
reasons) and are rescued. These include many avant-garde films, home
movies, amateur films, etc.

But any large archive has millions and millions of feet of film on
their shelves. There is simply no way that the bulk of that material
will ever be preserved -- even a single access print struck --
because there's neither time nor money to do so. It can cost an
archive (with their own lab) $50,000 or more to make preservation
elements of a single B&W 35mm feature film. Restoration expenses can
double that. That's preserving perhaps 10,000 feet of film. Do the
math -- there's not enough time or money to preserve (let alone
restore) most films.

And of course, that doesn't deal with the "mystery cans" of film
every archive has -- to really determine the contents of a roll of
negative one needs to print it or at least transfer it to video, a
large expense for an unknown quantity.

But treasures lie on these shelves -- a favorite recent example of
mine is the discovery, at the Library of Congress:

"Larry Appelbaum was thumbing through some old Voice of America
audiotapes about to be digitized at the Library of Congress when he
made a discovery that would stun him and many other jazz fans. Eight
10-inch reels of acetate tape were labeled "Carnegie Hall Jazz 1957."
One of the tape boxes had a handwritten note on the back that said
"T. Monk" with some song titles.
Appelbaum, a jazz specialist at the Library of Congress, got excited
at the prospect of finding unpublished materials by the jazz master
Thelonious Monk. Then he heard another distinctive sound. "I
recognized the tenor saxophone of John Coltrane and my heart started
to race," Appelbaum says. The Nov. 29, 1957, concert was recorded by
the Voice of America but never broadcast. For years, the recordings
were lost and forgotten. Now, thanks to Appelbaum's discovery, Blue
Note Records is releasing them."

Of course, it's relatively cheap to preserve audio (no DAT jokes,

So what to do about film?

First, a video transfer is NOT preservation -- at best it is an
access copy. A video transfer made on a device like an "Elmo" can do
more harm than good -- damaging shrunken film while making a very low-
quality video transfer.

There are fewer and fewer film labs around, and only a handful do
preservation and restoration work. Many are really good, but expensive.

Digital scanning holds promise, but most scanners are designed for
new, unshrunken, undamaged film, and are very slow -- some take
several seconds per frame. Most are designed for 35mm, with 16mm as
an afterthought, and other formats are rarely supported. Some
facilities price digital scanning at well over a dollar per frame,
even at 2K resolution. Add to that the cost of storing this data,
and migrating it to new media every few years... a mess.

A couple of years ago, I designed a special digital scanner for the
Library of Congress Motion Picture Conservation Center, for scanning
the Paper Print collection. These were films that had, for copyright
purposes, been printed (pre-1914) onto 35mm-wide paper (some
perforated, some not) and deposited at the US Copyright Office.
There had been many projects over the years designed to transfer
these films back to film. with varying degrees of success. Most
involved manually lining up and photographing each frame, a painful
process that produced jittery images very very slowly -- perhaps 3
frames a minute with a fast operator. Of course, this produced mere
copies -- dirt, dust, scratches, jitter, contrast problems, blotches,
etc. remained intact.

The scanner (part of a complete system) works at 2K resolution (more
than enough for these prints) and scans at speeds up to 16 frames per
second -- essentially realtime, for silent films. While designed for
paper, it works well with film, and is capable of handling extremely
damaged material -- shrunken, mis-perforated, torn-perfs, etc. --
without requiring extensive repairs first. (Note that repairing torn
perfs on shrunken film is extremely demanding and time-consuming, as
it's done with clear tape and an X-acto knife. A reel can take a
week or more, depending on degree of damage.)

We can currently scan 2K at 16 fps, and 4K at 5 fps -- and 1600 x
1200 (a good choice for regular 16mm) at 32 fps. This may change
radically in the near future, with very fast high resolution scanning
-- 24 fps or greater at 4K. (Note that most current Hollywood
effects work is done at 2K, so 4K is very high resolution, able to
preserve all the detail in 35mm material.)

Of course, there's the issue of data storage. Storing uncompressed
4K scans (at 12 bits, each frame would take over 18MB, or 27 gigs per
minute). But the second you are dealing with 4K scans, wavelet
compression -- JPEG 2000 or Cineform -- can be very helpful, and
visually lossless -- about 40MB/second, or 2.4 gigs per minute. On a
400 gig LTO-3 tape, costing $65, holds 166 minutes, or a cost of
forty cents per minute -- 2x that -- 80 cents -- for two copies. 2K
scans at 12 bit would normally take 5MB per frame, or nearly 7GB per
minute -- lightly compressed they take about 700MB/minute. That's
570 minutes on a $65 tape, or, if stored on two tapes (one a clone)
the cost is about 25 cents per minute.

Of course, LTO-3 tapes should be cloned every five years or so, so
that is a future expense (though the process can be done
automatically with very inexpensive autoloading tape drives.

In the near future, holographic disks are coming. The first version
of these will store 300 GB on a $150 disk, with an estimated storage
life of 50 years. More expensive, but less cloning required over the
years. Prices will drop quickly as these become more common.

But what does it cost to scan film? Using our most expensive
scanner, which is overkill for most users (especially on this list)
the cost, based on running it 40 hours per week x 48 weeks per year x
5 years, is $28/hour, plus operator.

If you are scanning at 2K, at 16 fps, you can scan about 35 minutes
per hour -- so the cost per minute would be less than $1/minute for
the machine, and another 25 cents for tape. Add another 50 cents to
make a couple of access DVDs of the scans (done easily on a cheap
computer from the same scans). So, $1.75 per minute.

If you are scanning at 1600 x 1200, you can scan about 70 minutes per
hour, so the cost drops, with tape, to under $1/minute (plus
operator, space, electricity, etc.).

Our less expensive scanners (which have some limitations, but the
quality is the same) will reduce this cost further.

If you take the $40,000 needed to make preservation elements of a
single B&W 35mm feature, you can scan and save 380 HOURS of film at
2K resolution.

4K is currently more expensive, because it's slower -- about 10
minutes of material scanned per hour -- so the cost is about $30 per
minute, plus another dollar to store the data on two tapes, and the
cost of DVD access copies. When faster sensors come along, it's a
simple upgrade, and costs will drop to below 2K levels.

Of course, this only preserves the work -- it doesn't restore it.
Actually, part of the work of restoration is included in the scanning
-- most scratches and dirt and dust that aren't printed into previous
generations of the film are minimized by a special lamphouse design,
and most film damage (except for open splices) doesn't need repair.

There is less expensive software coming that can do batch processing
(no operator required) of dirt/dust/scratch removal, deflickering,
and stabilization -- of course you still have your original unaltered
scans archived, so as software improves you can always go back to the
source if need be.

Color correction and grading can be done using inexpensive tools
these days, and digital versions (HD, 2K, etc.) can easily be
generated at little expense. If you want to go back to film, of
course it costs more, but we also make reasonably low-cost film
recorders that will help keep the cost down.

Not trying to promote anything (didn't even mention the company name)
but I am very excited about making preservation so affordable that
it's actually as cheap as re-canning!

We also are finishing a low-cost "desktop" scanner for film schools
-- HD or 2K scans directly compatible with Final Cut Pro -- shipping
this spring.

Jeff "got to get back to work" Kreines

For info on FrameWorks, contact Pip Chodorov at <email suppressed>.