Film restoration in the digital domain: A chat with James White

Maria Falconetti in Carl Theodor Dreyer's The Passion of Joan of Arc (1928). This is how the frame appears in the current Eureka!/Masters of Cinema Blu-ray disc edition of the film. Below, the image before restoration. 

I first heard from James White, the British film restoration maestro, a few years back, around the time the below-mentioned ITV Blu-ray of Powell and Pressburger's Black Narcissus came out in Britain. White and I reconnected around last year's Christmas holidays, just as two very disparate projects in which he had a restoration hand were coming on the scene: the above-depicted version of the Dreyer masterpiece, and Arrow Video's iteration of Lucio Fulci's notorious Zombi, known in Britain as Zombie Flesh Eaters. We figured the time was right to catch up on some of the issues concerning film restoration in the brave and still relatively new but increasingly pervasive world of digital, and thus an e-mail exchange began.  It took some time, but I think the results are informative, and contain news that is both heartening and distressing. If you can, you should read in conjunction with Nick Wrigley's essay "Crimes Against The Grain" in the December 2012 issue of Sight & Sound. Wrigley was one the founders of the Masters of Cinema imprint and worked as a restoration supervisor with White on the Dreyer project. The interview with White follows: 

1) Film restoration seems to have pretty definitively shifted from a photochemical process to a digital and electronic one. When was the tipping point for this, and what does it mean in terms of the economics of film restoration? Not to mention the distribution of restored motion pictures?

It's certainly true that over the past decade or so, film restoration has moved from a largely photochemical process to one relying almost completely on digital technology. When I began working in this field in New York back in the mid-1990s, digital restoration was still in its infancy and most of the major projects such as those being overseen by Photoplay (Napoleon) or Harris & Katz (Spartacus, Vertigo) were being produced almost solely through traditional photochemical processes. These days, however, if someone embarks on a new restoration, they will almost always begin by scanning the best existing film materials and working in digital throughout the entire process. 

My time at the BFI reflects this shift in approach. When I started working there in 2002, film restoration was still very much the preserve of the archive lab, with digital technology having little to no involvement. Restorations from that era represented the best that could have been achieved at the time, but in recent years the BFI has seen fit to revisit many of their key titles using digital tools, having seen the advantages these bring. A title like Blackmail for instance, restored last year by the BFI as part of their silent Hitchcock project, now looks amazingly improved in its digital incarnation in comparison to the prior restoration completed in the 1990s. 

I don’t think there’s been a single tipping point that made this change happen, other than restoration simply mirroring the film industry at large, and its shift to using digital tools for pretty much everything under the sun. The simplest reason though, is that the tools just got better. Most importantly, the software became sympathetic to the needs of archive films and offered solutions to problems that traditional restoration hadn’t ever been able to deal with sufficiently. Issues that had always been difficult to impossible to fix - things like image stability, density issues, deep emulsion scratches, warping, registration issues, deterioration brought on by heat or moisture, and damaged or even missing frames – these could now be improved upon in a way that just couldn't be done by traditional means. 

Having access to these new tools hasn't made the work cheaper, though. If anything, restoring a film has become more expensive simply because audience expectations are so much higher now. Fortunately there are so many more outlets to distribute these titles, so the economics are there to support the work. Mind you, there's definitely a limit to what most distributors can afford. A top-tier restoration like The Red Shoes or Lawrence of Arabia can demonstrate beautifully what digital restoration is capable of - these films now truly take your breath away - but the vast majority of film titles don't enjoy anything near to the restoration budgets these projects had. So while restoring the majority of film titles to a 2K/4K level might be beyond the means of some distributors, restoring a film to a very high standard in HD is often within their means. I should mention that going this route doesn't support a film's archival basis - restricting your end result to HD doesn't create what most archivists would agree is the basis for film preservation - but it does provide distributors with a format suitable for the majority of current release platforms (DCP, Blu-ray/DVD, etc). 

In any case, what's encouraging is the high amount of restoration work currently going on internationally. Rights-holders of film libraries have started to see the value in restoring their titles to make them accessible for HD broadcast and online platforms. DVD/Blu-ray sales have helped create new audiences for archive titles, and the improved presentation quality has made viewers more aware and appreciative of the practice of film restoration on the whole. Back when I started, Criterion was the only company around that devoted time, attention and a decent budget to giving a film the best presentation possible on video, but now in addition to well-established labels like Eureka/Masters of Cinema, the BFI, Milestone and Kino/Lorber, you've got all these new labels like Olive in the US, Second Run in the UK and Edition Filmmuseum in Germany resurrecting lost cinema or rarely-seen classics.  It's the work of companies like these that is really helping to bring new interest in film restoration. I mean, it's great when Casablanca or Singin' In The Rain gets treated to another state-of-the-art restoration, but there now seems to be an actively growing interest in discovering new films and film collections from all corners of the globe. A project like Milestone using Kickstarter to fund a new restoration of Shirley Clarke's Portrait of Jason is a great example of the public taking an active role in film restoration. Likewise for Distribpix's recent restoration of The Opening of Misty Beethoven - that film now looks better than anyone probably ever expected it to!  

2) The tools for film restoration have evolved to the point that the lay viewer is under the impression that it can perform miracles, which to some extent seems objectively true. So let's look at it from the other end of the telescope, as it were, and talk about the things that the tools CAN'T do.

Well, as I said there’s a great amount of things that digital tools can do to repair the issues that have always blighted older film titles that we couldn’t do just a few years ago. That said, it’s important to bear in mind that in the most fundamental sense, no amount of restoration, digital or otherwise, can significantly improve the quality of the image of the film element that it’s sourced from. The issues I’ve mentioned - dirt, scratches, stability, flicker, missing or damaged frames, etc – these can all be improved significantly through the careful application of digital tools. But the basic details within that original image, meaning the film grain, the level of detail, etc – these can only be improved within the parameters of what the original element would allow.

It’s like up-rezing an image from SD to HD. An increase in detail doesn’t happen simply because you’ve added more pixels. Of course one can give the illusion of increased detail or sharpness through digital enhancement, but then it’s no longer the film you’re working with, it’s something else. I work to create the best representation of an old film possible, but that means keeping the results within the historic bounds of what that film would be able to achieve at the time – be it its film grain, the saturation of its colors, the level of its contrast, etc. A film from 1930 shouldn’t look like a film from 1950, and it most definitely shouldn’t resemble a film from 2013. Whatever tools you’re using should be done in service of what the film looked like at the time of release, not some ill-advised notion of what it could look like now. That, for me, is probably the most important thing.

Unrealistic expectations are a common source of frustration for film restorers, as often the best existing film elements simply won’t allow for a "miraculous" result.  The project is only as good as the material it's sourced from, and not all films have been subjected to the same treatment over the years. Often with older titles the original negative has been lost, or is far too damaged to work from. In this case you might be relying on the next best thing, like a second-generation interpositive or fine grain element. But there may be problems with these materials as well. Often restorations have to make use of more than one element, each subjected to vastly different histories, and then it can prove a real challenge to make the overall results consistent. To put it simply, every film is different, and every restoration has its own set of challenges.

One interesting development of late has been the discussion of the differences between what we’re now able to see digitally in comparison to what was visible on a film print. The Wizard of Oz scenario is well-known – in Warner Brothers' new Technicolor restoration you could suddenly see wires holding up the Scarecrow, whereas in all previous releases of the film, they were invisible. Clearly back in 1938 Victor Fleming and his Director of Photography understood enough about the photochemical process that they could expect the wires to be invisible by the time theatrical prints were created in the printing chain, but in working digitally from the first generation elements, the wires are now there, clear as day. Likewise, I've heard it recently remarked that with Vistavision films such as Vertigo we’re now seeing a level of detail on Blu-ray that was never intended for the cinema, as theatrical prints would have been made from separate matrices reduced for printing. So how does one approach situations like these? It’s an interesting dilemma.

3) What are your favorite/favored tools, or the ones you find yourself applying more often? What have you had a hand in developing? 

I can’t claim to have had a personal hand in the development of any specific software tool, but I’m fortunate that I work with such an amazing team of technicians at Deluxe Labs in London, who are always working to find new ways of meeting the numerous challenges this kind of work presents. We’ve worked together on so many projects over the years for the BFI, Eureka and most recently Arrow (Zombie Flesh Eaters) we’ve developed a great rapport together, which is crucial when you're working on archive titles. 

With picture restoration there's a variety of tools at hand that benefit some films more than others. I'm always cautious about any so-called "automated" processes, as it's so easy for these tools to have an unwanted effect on film grain and detail even if applied carefully. So while it takes far longer to do, nothing works better for me than old-fashioned frame-by-frame cleanup, performed manually. It can be incredibly complicated, time-consuming work, but there's just no substitute for treating each individual frame with the utmost care. Full disclosure, though - I don't do any of the hand-on stuff myself anymore. As a restoration supervisor, I oversee all the steps of a project from the initial research and selection of film materials to the reviewing of scans to the full grading of the film to all stages of picture and sound restoration. Whenever possible we'll work with the director or director of photography (if alive and available) and we always work to deliver the highest quality representation of a film's original release within the parameters the budget and time frame allow for.   

4) The two projects that were the springboards for this correspondence were pictures that seem to have almost as wide an aesthetic gulf as a temporal one: Carl Theodor Dreyer's 1928 masterwork The Passion Of Joan Of Arc, which you supervised the Eureka!/Masters Of Cinema Blu-ray version of, and the aforementioned Zombie Flesh Eaters, the notorious 1979 Lucio Fulci horror cult fave. In terms of materials, each one must have presented unique challenges. But there's also a question of properly serving a given film's aesthetic. I understand I'm constructing a bit of a big tent question here but I'm interested in your thoughts on all of these considerations. 

One could hardly choose two more different films than those two, could they? Though I still think they'd make a great double bill! Hey, they both feature heavy doses of pain and suffering, both base the reasons for their violence in religion of one kind or another, and both really know the power of a well-executed close-up. Although it's true that only one of the films features a punch-up between a zombie and a shark...

Anyway, the fundamental criteria for me is always to present the film as close to how it originally appeared in the cinema. That means restoring the film to the highest quality possible but not employing any means to "update" or "improve" the image or the sound in some misguided effort to refashion it to fit in better with modern-day expectations. I want the results of something I restore to appear as a film artifact, not a digital one, which is why I've been using the term "preserving a film's photochemical integrity" lately. 

The Passion of Joan of Arc project grew out of Eureka's interest in releasing the film on Blu-ray and the Danish Film Institute's desire to see Carl Dreyer's original Danish version made available to the public. As far as I'm aware, Gaumont are still planning to restore it themselves, but the version they eventually produce will almost surely be the French version we've all been familiar with for years. So as this stood as the only likely chance to have Dreyer's debut version seen to properly, we wanted to make sure we presented Joan in the best and most accurate representation possible. 

We had the good fortune to be working from the Danish Film Institute's preservation materials, struck directly from a first generation 35mm print discovered in Oslo. This is as literally good as it gets with this film, a film we're amazingly lucky still exists in any form given the problems it faced from the very beginning. For those unfamiliar with the saga this film was put through, the story bears repeating, so I'll crib a bit from a piece written about the restoration for Moviemail last year:

"The trouble began just six months following the film's Copenhagen premiere in 1928 when the original negative was destroyed in a fire and the two original prints created from this element were subsequently lost. A new second negative, incorporating extra footage not featured in the original version, was subsequently cut together by Dreyer but this material swiftly fell victim to misfortune, and was presumed to have perished in a lab fire as well. Although fires like this were fairly commonplace at a time when highly flammable nitrate film stocks were used, the fact that the film had been destroyed on two separate occasions seemed to imply that The Passion of Joan of Arc was destined not to survive in any form.

Over the years that followed, several incomplete prints and material believed to represent the second lost negative re-surfaced; as a result numerous versions of the film have been edited together and screened for international audiences, but Dreyer’s original version remained lost, presumably forever. 

Then in 1981, one of the original prints struck from Dreyer’s first negative was miraculously discovered at a Norwegian psychiatric hospital. The Danish Film Institute (DFI) immediately acquired this material, which came to be known as the "Oslo print" and created a new preservation negative, guaranteeing a secure future for the film and keeping the elements in optimal conditions for years to come. Finally, after over half a century, The Passion of Joan of Arc could finally be seen exactly as Dreyer had originally intended."

Before-and-after restoration images from Joan.

So with all that history facing you, the responsibility you bear to remain faithful to the film and the materials at hand should be always first and foremost on your mind. Given that we had a specific budget and fairly tight schedule to work by, my approach was to do the best job possible but to be wary of not "over-restoring" anything. The entire film required stabilization due to the shrinkage and sprocket wear the print had endured throughout its 80-plus year lifespan. Every join between shots required manually re-setting as there were bumps on literally every cut - a fairly daunting task for a film as creatively and heavily edited as Joan. Density inconsistencies manifested in the form of heavy flickering had to be significantly reduced. There were numerous instances of damaged and missing frames, often requiring new frames to be digitally interpolated, as well as a steady stream of heavy and light scratches throughout. Not all of these issues could be remedied completely, and some problems could only be marginally improved. If you watch the restored version of Joan, the film still exhibits quite a few of these issues, although it's now a fraction of what we started with. The important thing is that none of the work we did compromised the original photochemical look and feel of the film, and the original texture of Joan remains. Fortunately most of the feedback we've received on the Blu-ray has been very positive, so I think on the whole people appreciate this approach, even if the results aren't 100% pristine, which would be a very tall order with any film from this era. 

Zombie Flesh Eaters was quite a bit different, in that it's a film that's been widely available for years, most recently in a fairly decent Blu-ray presentation from Blue Underground. But Arrow felt strongly that the film deserved better, so they decided to fund a new restoration of the film on the basis that we would be granted access to the original Techniscope negatives. A little digging revealed that these elements hadn't been made available for transfer since the film's original print run (contrary to a good deal of misinformation that's been circulating on the web), so we had a good opportunity to correct for some home-video crimes of the past and restore the film to its original release presentation. 

Techniscope was a fairly popular format in Europe throughout the 1960s and 70s - it was essentially a low-budget version of Cinemascope, using two perforations instead of the standard four. After the initial negative is processed and edited, secondary elements (Interpositive, Internegative, etc) are produced by blowing up the image to fill the 4-perf frame. Having worked on another Techniscope project a couple years ago for the BFI (La Vallée, Barbet Schroeder, 1972) I knew that using the negatives provided the ability to capture a good deal more image area in the film frame than any of the other printing elements in the film chain, as the blow-up process would have forced the cropping of all sides of the image in order to fit within the 4-perf frame. So working from 2K scans of these original negatives, I was able to bring the color palette, the grain replication and all the details present on the negative back to its proper place, as well as reveal a good deal more picture area on all sides of the frame for most of the film. None of the work was exceptionally challenging, although the footage from the famed shark-zombie sequence was blighted by a lab fault that proved impossible to do anything but digitally minimize - fortunately this issue has always been a part of the film, as it was printed into every version anyone's ever seen! 

The reviews we've received for Zombie have been overwhelmingly positive - considering how it was represented on VHS here in Britain for so many years, I think people were amazed to see the film looking as good as it did! That said, It shouldn't really surprise anyone familiar with Fulci's work - he's a director with an amazing eye, the camerawork by Sergio Salvati is frequently stunning, and the combined use of locations, colour, and music by Fabio Frizzi make Zombie one of the best films of its kind. That said, it does bring up something interesting in terms of your question about properly serving a given film's aesthetic. Might some viewers actually prefer the rough VHS-era representation they grew up with, complete with faded colour, horrible sound, video snow and tape damage? It certainly makes for a very different viewing experience, but one I suppose that should be treated as just as valid, as it was viewed by so many people, particularly in this country, in exactly this way.  

So in a nutshell, my criterion is to represent the film in as close an approximation to how it originally appeared on screens as possible. More than being just a technical preference though, I view it as a true responsibility that I wish everyone would take very seriously. Which is why I'm disappointed when Gaumont Pathe chooses to apply a headache-inducing level of de-graining to Les Enfants du Paradis. Or when Hammer chooses to replace the original optical effects in The Devil Rides Out with new, "modern-day" digital versions. Or when so many recent restorations of classic films coming out of Italy are digitally sharpened to the point of resembling a Playstation game. Or when Pathe choose to put Le Samorai through so much image processing that each frame of Alain Delon in his raincoat and fedora end up resembling an oil painting more than anything captured on film. These are extreme examples, but there are plenty of others. Given the expense of restoring a film, practices like these should be actively discouraged, as the results will likely be the only means of seeing that particular film title for years to come. Considering the fact that the use of celluloid is quickly vanishing as we speak, and that very soon simply being able to project 35mm will be a thing of the past, it's essential that the tangible, textural look of celluloid is preserved properly and accurately. 

 5) Back in the late 1990s, I was at an event in New York City at which the great cinematographer Jack Cardiff spoke. In the audience was Robert A. Harris, who asked him if the lavender tint on Black Narcissus was deliberate. Cardiff's answer was something like "What lavender tint?" and hence a great error in film preservation/presentation started on its way to being corrected. I believe we first corresponded after the ITV restoration of Narcissus that you were involved in, a project that occurred even after the removal of the lavender. My question is, how important is the input of someone involved in the original work, and in the event that no survivors are around to supply input, what are the research roads you and your team find yourselves most frequently traveling?

Well first I should make it clear that I wasn't part of the restoration team responsible for Black Narcissus, although I was there in the periphery when it was happening. That restoration was produced by the archive team at ITV, with Thelma Schoonmaker and Jack Cardiff supervising throughout the process. I was fortunate enough to be invited to some work-in-progress screenings in which different film stocks were being tested to replicate the original 3-strip Technicolor look of the film. Thinking back, this served as a pretty good illustration at the time of the limitations of modern film stocks and their ability to replicate the look of older formats accurately, and the help that digital, if used with care, could provide in this regard.  I was lucky enough to meet with Jack Cardiff off and on during my years at the BFI, and true gent that he was, he was always happy to answer the numerous questions I was constantly springing on him (and I certainly wasn’t the only one). I've also had the great pleasure of working with Thelma Schoonmaker, most recently on BFI's Blu-ray restoration of what Michael Powell considered his first "true" film, The Edge of the World - an amazing film, if you haven't seen it.

Anyway, there's no real substitute for having the original talent in the suite with you while you're working on these films. Unfortunately when working on older films this just isn't possible a lot of the time, so it's imperative that one has a real understanding of a film's history and the way it's supposed to look and sound. One of the alarming aspects of the expanse of digital technology is that more and more people entering this industry will have had virtually no hands-on film experience. How is one supposed to accurately restore a film if one doesn't understand the photochemical properties of the medium? Small decisions can have an enormous impact on a way of film is represented, and those decisions need to be informed ones.  

I've already mentioned the importance of maintaining the original film grain structure, but grading the image correctly is just as important. From Eastmancolor to 2-strip and 3-strip Technicolor way back to Dufaycolor or the experiments of Friese-Greene, there's been so many ways to interpret color throughout the history of cinema that one needs to be very careful in how those colors are represented in any new restoration.  The temptation to make the colors pop, to over-saturate, to make them much deeper and heavier than they would have looked on an original film print, is one that needs to be kept in check. Working from negatives, especially those that haven't been subjected to much fading over time, gives one a great deal of latitude when grading, and as a consequence the films can be as graded as warm or cool as one likes, so one needs to approach the grading with care. It's the same story with black and white titles. There's a tendency by some people to push the contrast on their black-and-white titles a bit too heavily, in an effort to make the image appear bolder and sharper, thereby giving it a look possibly more dramatic to the modern eye. But going this route is often incorrect, as most film print stocks from earlier years wouldn't have been able to support this look. As a consequence the tones in the grayscale often get lost and the subtleties of the original light and shade are simply "graded out" in favor of heavier contrast. Fortunately, the adoption of HD and Blu-ray have brought improvements to this area, as the higher resolution allows for proper replication of these subtleties. 

My thoughts about audio are pretty much the same, meaning I'm really not in favor of presenting any film in an audio format not from its time. So 5.1 or 7.1 mixes for film originally released in mono I realize it's done as an effort to make young audiences more likely to accept these older films presented in this way, but it's historically irresponsible and about as convincing to the ear as "electronically re-channeled for stereo" was in the 1960s. 

I'll give you an example of the range if choices you can be presented with. A few years ago, I was working on Terence Davies' The Long Day Closes (1992). The grading operator and I had a day to get things ready before Terence Davies and his DP Michael Coulter arrived, so we spent part of that afternoon applying an overall grade to the first couple reels. Right away we could see that the colors held in the negative were so bold and rich, that if we notched up the saturation a bit it looked just like a Douglas Sirk film! So for fun we graded the film as if it were, with bright and lurid primaries on show throughout. Now as anyone who's familiar with the films of Davies knows, he's extremely specific when it comes to color and has often used it in a creative way to evoke specific feelings for the past. So as soon as he and Coulter arrive, the first thing they tell us to do is drain all the color from the image and to start again. As with Distant Voices, Still Lives (1988), Davies and Coulter had applied a bleach-bypass to The Long Day Closes giving it a carefully faded, almost sepia feel throughout. As the material we were working from hadn't been subjected to this lab process (this is traditionally done later on in the printing ) we could have very easily graded the film in any number of ways, including our attempt at 1950s melodrama. Of course we wouldn't have done this as we were well familiar with the film and it's printing history, but one could easily imagine a scenario in which the person restoring the film hasn't done his homework, or doesn't have the filmmakers on hand to show him the way, and you'd wind up with a version of the film that would be completely wrong, both historically and aesthetically. 

That said, there are times when one can be too deferential when working with original talent. I'm talking about those rare occasions when a director has decided this is his opportunity to "fix" an aspect of the film he was never satisfied with, or "update" it in some way to bring into line with his current thinking. The results can often be interesting, but producing some updated version should never be done at the expense of presenting the film as it was originally released. I don't have a problem with Lucas or Friedkin or Coppola re-working their film every so often provided they allow the original films to survive intact. It's when this doesn't happen that I feel there's been an abuse of power, as somebody's personal agenda has created a deliberate distortion of a film's history. I've never subscribed to the idea that the director or any one person "owns" a film - a film belongs to history, and it's our responsibility to ensure that that history is accurately represented.

UPDATE: James White writes in, via email: "To answer Mr Kroll's question in your comments section, how these restorations are archived really comes down to the level at which they're restored. The current industry spec for archiving file-based film work is to store them on LTO5 archive tapes, which, as per good archive practice, should be cloned and the copies stored separately. If a restoration is being produced to the level of 2K/4K result, then the resolution would be high enough to justify producing a new Digital Internegative (DI), which was until recently the source for creating new restoration prints. Truth be told, there's still no better-proven long-term storage medium better than film, but sadly this seems to be happening less and less - a clear consequence of the diminishing need for film prints and the drive to go file-based for as many things as possible. That said, one should take heart that restoring a film anew would never result in the original film elements being discarded. On the contrary, the availability of a new restoration, at whatever level its been produced, would likely mean that the film elements would then be left alone and safely preserved for years to come. "