Monday, November 29, 2021

Reconstructing a "Lost" Version of The Phantom of the Opera in HD

I sometimes like to think of fan-based restorations of classic films, whether they are of a Godzilla film, a Walt Disney classic short or Star Wars, as the culmination of film restoration.  The Internet and broadband has fueled talented creators to take films from whatever sources they can find and make something even better than the studios which made or have rights to these films did or could.  I have talked about this subject before and in the recent past I have done a few minor projects myself.  Unfortunately due to the nature of the films such projects often cannot be shared publicly.  My most recent project is safe enough to share and touches upon a subject I have discussed in the past, the 1925 version of The Phantom of the Opera.  Specifically my goal was recreate the first version of Phantom I ever saw, the Killiam Film Classic release.  Sit back and let's go over why this release was significant, how it entered my notice and how I restored it to what I hope will be seen as its "proper glory."

Part I – Silent Film Comes to Television

Back in the late 1950s, a comedic entertainer named Paul Killiam from New York had the idea of delivering humorous commentary to silent films, somewhat like an early version of Mystery Science Theater 3000.  In this period, silent films were, for the most part, easy to find and often in the public domain because the studios never bothered to renew the copyright to these films after their 28th year.  Killiam had some success with his comedic commentary but he was also a genuine enthusiast of silent film and helped to preserve many films.  He broadcast digest versions of many of them with narration in the TV series “Silents, Please!” from 1960-62.  During this time network affiliates and independent stations were looking to fill airtime with inexpensively produced content and offerings like "Silents, Please" were popular and fit the producers' needs.  

As the 1960s progressed, sound films became much more commonly run on TV and calls for silent output from Killiam and others were becoming . . . quieter.  He packaged a series of films, including Phantom, for a series called “The Silent Years” for more highbrow TV networks like NET and PBS.  Included in this package was a complete version of The Phantom of the Opera, which had opening and closing title cards identifying it as a “Killiam Film Classic”.  

For this version, Killiam had Karl Malkames restore and tint a 35mm copy of the Eastman House Print with a good degree of faithfulness to the original tinting continuity as the tint colors were known in the late 1960s.  Killiam also commissioned distinguished silent film accompanist Gaylord Carter to compose an organ score specifically to accompany his new version.  This version was copyrighted in 1971 and played on US Television for at least two decades.  I do not know if other versions routinely played on TV, but this was the method by which the Phantom was introduced to a new generation of film and TV watchers.  

Part II – In Which a Young GH Encounters the Opera Ghost

My own personal history with The Phantom of the Opera began with the Killiam Film Classic version.  I encountered it on A&E, when A&E could still plausibly call itself the “Arts and Entertainment Network” in the late 1980s.  I taped it off cable and watched it frequently for years.  For many years after this tape was the only silent film tinting and scoring I had any experience with.  (I had a Goodtimes Home Video version of Nosferatu but it was in pure B&W and silent.)  These were the days before “speed correcting” silent films for video was a common practice, so the action was sped up to run at 24fps when the film was intended for projection at 20fps.  Killiam did perform a bit of "mechanical" speed correction on his version, as I will discuss later.

When I was older and had the funds to buy VHS and DVDs, I wanted to replace my old VHS off-the-air taped copy of Phantom.  That VHS tape went to the landfill at some point, which modern-day me would have been happy to transfer and upload for the original commercials.  Unfortunately for years I could not find the version with the same score or tinting as the Killiam Film Classic version.  As far as I can tell, it has never been released on home video (except by some bootleg outfits maybe).  Even finding a complete copy of this version was rather difficult.  Some years ago I found a copy on YouTube that had the first few minutes of the Killiam Film Classic version and later another YouTube video which had all but the opening Killiam title cards.  

In the US, with two exceptions, all quality disc releases of the Eastman House Print have been done through and David Shepard's Blackhawk Films.  Gaylord Carter did another organ score for Blackhawk’s version which was recorded in 1974.  I did not know of this second Carter score until I bought the Image Blu-ray back in 2012 or so.  Carter’s Blackhawk score has similarities to his Killiam score, but they are quite different.  It is indisputable that Carter's score for Killiam came first, having copyrighted in registered 1972 and perhaps first shown on PBS as part of the second series of "The Silent Years" in 1975.  Blackhawk did not advertise its Carter/Phantom score until October of 1976.

I did not know when I was younger that during the 60s and 70s there was a market for 8mm, Super 8mm and 16mm film prints for home users with the appropriate projector.  I have since learned that Blackhawk offered its version of Phantom on these formats during that period and eventually on VHS and Laserdisc by 1990. I never had a film projector or a laserdisc player, so any encounter I may have had with a tinted and scored version of Phantom was either via my taped copy of the Killiam Film Classic release or had to wait until the Internet made video releases easily available.

In the Blu-ray releases from Image and later Kino, Shepard used an HD scan of the Killiam Film Classic print for the video and used Carter’s Blackhawk score as one of the audio options.  Shepard acknowledged that the synchronization between the music and the action comes and goes a bit because Carter synced his score to a slightly different version of the film.  Carter’s Killiam score was nowhere to be found because Blackhawk never had the rights to that score.

Part III – Recreating the B&W Bal Masque Sequence in Higher Quality

Much more recently, I was able to find a YouTube Upload of the Killiam Film Classic version which had much better quality than the previous uploads.  The resulting video was still SD quality and the audio noisy, but it was better than some quality DVD releases.  I grabbed it using 4K Video Downloader and found that the program somehow managed to get to me in progressive scan.  Fortunately because the Killiam Film Classic version runs at 23.976/24fps, it can be readily deinterlaced through a standard 3:2 inverse telecine algorithm.  Silent films speed corrected from VHS even into the Blu-ray era used uncommon frame interlacing methods to get a film which may have run at 18fps or 22fps to 29.97fps.

While I was still busy working on an earlier project to tint and add different soundtrack to the BFI Blu-ray's HD scan of the Hampton Print, it stuck me that the shots and angles used in the Bal Masque sequence of that Print matched that of the Eastman House Print.  The Eastman House Print, as it sits in the archives of the George Eastman House, has the B&W version of the Bal Masque sequence.  The style of the intertitle cards is different between the Hampton and Eastman House Prints, but the live action is the same (or so I thought).

I observed that the BFI’s HD scan of the Hampton Print had higher quality live action of the B&W Bal Masque sequence than anything else I had seen.  16mm scanned in HD can sometimes beat 35mm scanned in SD.  So with a decent source element on my hard drive of the B&W Bal Masque sequence as it appeared on the Killiam Film Classic version, I decided to see if could recreate it using the HD footage from the BFI and the SD titles and a few bits of live action from the YouTube upload run through a grayscale filter.  Much later on I discovered the Unsharp Mask filter and applied to reduce the soft look of this sequence.  

The end result was satisfactory, so from there I figured that recreating the entire Killiam Film Classic version would be easy.  Of course I figured wrong.  My goal was to marry the much higher quality HD video I had of 95% of the film to the soundtrack with tints that came close to the Killiam Film Classic and good sync between the film and the soundtrack.

Part IV – Recreating the Video of the Killiam Version

Tinting the film was harder than it first appeared.  I knew the tinting choices Malkames made well enough, although without a visual reference I probably would have made more mistakes than I initially did.  I encountered issues with trying to select a tint that came close to matching the original tints without doing too much damage to the fine detail of the HD footage.  His blue had a muted look most of the time, but during the Opera Roof scene it is quite vivid.  His green had a yellowish tint which I tried to preserve.  The video reference’s green was greatly oversaturated, but I do not know whether that was the original print or something which came later.

By my count, the tinting changes almost eighty times during the course of the film.  Sometimes one tint can cover several scenes, at other times the tint can change rapidly due to cross-cutting.  After so many errors and a rage-inducing VirtualDub 2 bug where the tint would change one frame before it should, I finally decided to cut up the film into separate files whenever the tint changed.  I took the opening and end title cards of the Killiam version from the YouTube Upload, and tried to sharpen the text and reduce noise, but I thought the results looked too artificial and left them as-is.

My next obstacle was to recreate the stretch printing which occurred in the Killiam Film Classic version.  I mentioned speed correction above, and there are sections of Killiam’s which use a kind of speed correction called “step printing”, where certain frames are duplicated as the film is copied.  Killiam used stretch printing for all three of the ballet scenes in the early part of the film, when the chandelier comes crashing down and during the Phantom chase.  The HD source of the Eastman House Print I used came by way of Blackhawk, which only stretch printed the first Ballet sequence and the Phantom chase.  I believe Killiam did this because these scenes, when played back at an un-speed corrected 24fps, looked ridiculous even by the standards of his day.  I used AviSynth to duplicate every 2nd frame once to more-or-less match the frame count from the YouTube upload.

The final issue with video I encountered was that Killiam’s version used additional fadeins and fadeouts than which were included on the Kino HD master which I used.  I had serious problems with AviSynth’s fade filters with VirtualDub 2, so for most of the fades that needed to be added I added them a frame at a time, tweaking the brightness for each frame either linearly or exponentially depending on how the YouTube Upload looked.  I added a fadeout/fadein for the first shot of the Opera Exterior and Opera Staircase, a fadein/fadeout when Christine enters the Phantom’s lair and fadeins for the Bal Masque’s first live action shot and the High Above Paris on the Roof of the Opera title card.  These were all the extra fades I could find.  

Part V – Syncing the Carter Score

Then I came to sound sync, which turned into a more laborious process than I expected.  Once I had created a video file containing all the tinting, step printing and fades I needed, I found my version was about 15 seconds shorter that my YouTube Upload reference.  The first thing I discovered is that the footage with the Man with a Lantern is significantly shorter in the Killiam version than the Blackhawk version.  Once I trimmed him down and added a few black frames in scene transitions, deleted some black frames in other transitions and killed some stray live action repeated frames, I was able to get the music within 10 frames of the YouTube Upload.  I decided that 10 frames, which is almost .5 seconds at sound speed, was essentially unnoticeable in terms of sync for a silent film.

But then I came to a bigger challenge when I realized my B&W Masked Ball sequence was almost 300 hundred frames shorter than the equivalent sequence from the YouTube Upload.  This was even after I had added in one shot and part of another from the YouTube Upload to restore shots not found in the Hampton Print.  It turned out that the Hampton Print was missing a few frames from the beginning and end of most shots compared to the Eastman House B&W sequence, and those missing frames added up over time.

I thought of adding the missing frames from the YouTube Upload, but decided against it because the shifts in quality would become distracting and it would have been a lot of work.  I also thought of using the color masked ball sequence, but that sequence is even shorter than my B&W reconstruction.  Rather than butchering the soundtrack with my editing, I used Audacity’s change speed and pitch shift functions to get the music to fit within the sequence.  On the plus side, I decided not to crop the left and right sides of the image of the reconstruction to match that of the Eastman House Print.  The Hampton Print was transferred more-or-less with a proper 1.33:1 silent aspect ratio while the Eastman House Print was transferred with sound equipment, giving a 1.2:1-like aspect ratio by cropping the left side of the image where the optical track would be on a sound print.  

Once I had done the pitch shifting and speed changes to that section of the Carter-Killiam soundtrack, I was good with sync up until the Roof of the Opera scene.  Thereafter sync began to suffer again, and it was hard to track down the cause because the two versions are substantially the same hereafter.  The two differ only slightly again, an extra repeated frame here, a few extra frames of an intertitle there.  While these differences were not substantial they add up over time.  I did discover that the Killiam version plays the end credits for almost 60 frames longer than the Blackhawk version, but by that time I could afford to be more liberal with sync.  But the film still continued to perplex me from the time the Phantom has an encounter with Florine until he is trapped by the mob.

The Carter-Killiam score is very difficult to edit unobtrusively for a novice like myself.  It has very few instances of true silence where a note is not being played, sustained or released.  I finally did find two instances of silence which, when condensed by about 2.5 seconds, finally got the sync in the second half to where it needed to be.  I did look into noise reduction but I feared that further tinkering with the soundtrack would have lost more than what would be gained.

Part VI – Special Thanks

All this work would not have turned out nearly as well, video-quality wise, without materials provided by the user known as ElectricTriangle on the Original Trilogy forums.  Originally I was going to extract the video from my Kino and BFI Blu-rays, but I kept reading his notes to his own work on the Phantom and kept seeing the flaws in the video of those releases.  I asked him for permission to use his improved video from his “The Phantom of the Opera (1929) Rare Scores Collection” release as a basis for my reconstruction and he gave me even better B&W material he used with his “The Phantom of the Opera (1929) - Roy Budd Score v2”.  So I cannot thank him enough for all his assistance with this project encourage interested persons to take an interest in his work.

Part VII - Where can I Watch this?

I have included a link to a YouTube Upload so you can judge for yourself whether I came close enough to the original Killiam Film Classic version.  I have peppered this blog entry with images from my reconstruction.


  1. Thanks for all you do. I read and re-read your blog constantly and this was so exciting to read through. I’ll be watching this for sure.