Out of the Archive - 'The Informer': Restoring an Early British Sound Film

Where do film archives get their material? How should they catalogue it? How do film and video deteriorate, and how can they be restored? Out of the Archive will answer these and other questions by tackling different archival themes, accompanied by expert comments by archivists on the issues raised. In this first edition of Out of the Archive Andrea Kalas, Kieron Webb and Richard Yeoman-Clark discuss the BFI National Archive's restoration of The Informer (1929).

What makes The Informer technically interesting?

When The Informer was released in 1929, the transition from silent to sound film was taking place. The two technologies were both extant in cinemas, so two versions were made: one silent and the other with synchronised music and effects, and dialogue sections. The restoration of this sound version illustrates technically this fascinating time in film history.

The BFI National Archive worked on the restoration from October 2004 to December 2005. A restoration is always a group effort. Those most directly responsible for the technical work included Kieron Webb, who oversaw all of the technical details and shepherded the work through; Richard Yeoman-Clark, who restored the sound; Angelo Lucatello, who inspected the film materials; and Ben Thompson, who oversaw the quality of the grading and image.

Silent film actors had no language barriers, so this Irish story had Swedish and Austro-Hungarian nationals as its stars. Both were international stars, but were realizing their heavy accents would be a liability on the new sound stages being built in Los Angeles. Perhaps the slower conversion of British theatres meant their stardom was attractive to an industry that still saw a market for silent films. Like many films of this time the sound version of a film originally intended to be silent was a bit of a compromise. Actors were hired to speak their lines for the two stars. Both voices sound RADA-trained, not authentically Irish. This "ghost voicing" was noticed by reviewers. Most likely the voices were recorded by the actors just off-set while the stars mouthed the lines. A lack of lip sync was impossible to avoid. Many times the acoustic sounds wrong and does not match the picture. And, as the "ghost voices" are fairly close to the recording microphone, their dynamics and sound do not fit the on-screen actor.

The film wasn't "100% talkie". Only the final four reels had all dialogue, music and effects. The first five had music and effects only. The sound was probably captured in different ways but none of them professional by modern standards. We noticed instantly that many effects, like gunshots, appeared completely out of sync. It was most noticeable in the scene in which a needle is being placed on a phonograph. A scene in a train station when a truck is outside of the station featured the sound of the idling truck engine in one shot, but it disappears after the cutaway to the interior.

Sounds were recorded before the introduction of sound stages, and on discs rather than magnetic tape (during restoration, sounds indicating the lifting or dropping of a recording stylus were picked up). We assume the sound effects were done in an early 'foley' style – shots being fired while someone screened the film. We assume the music was recorded in a venue meant for performance rather than recording. For the film’s last sequence in the church, the recording of the organ was most likely recorded – in a church! The combination of primitive recording equipment and echoey acoustics results in distorted sounds.

However the sound was initially recorded, the final sound was captured using RCA Photophone, an early sound-on-film process still in competition to become a standard.

How did we go about restoration?

Restoration involves examining surviving copies of a film to find the best 'source material', and then identifying the technical steps necessary to improve upon it when making new copies. This is to correct either technical faults in the original film or problems caused by later deterioration.

Like all feature films of the time, The Informer was produced on nitrate film stock, which is unstable and inflammable. The Archive holds original nitrate picture and sound elements, but both (especially the sound) have suffered severe decomposition. However the Archive had made new masters on ‘safety’ film stock. These were chosen as the sources for the restorations. The basis for the restoration of the track was the Archive’s sound preservation master. Similarly, the only source used for image restoration was a picture element known as 'fine grain'. This is a copy of a film produced on a particular kind of film stock not suitable for screening, but ideal for printing further copies from.

Based on examination of these two elements, deductions and recommendations were made.

What problems were we faced with?

Both the sound and silent versions were 9 reels long. The sound version had music and effects only for the first five reels. Dialogue was represented by intertitles, until the last four reels. The same picture negative was used for reels 1 – 5 of both versions. The last four reels share some takes.

The sound track was not 'mixed', like films today. It was physically joined and 'blooped' at each edit. As a result of the apparent inability to fully mix and re-record, sound levels vary greatly both within and between reels. The bloops cause particular problems as just removing the thumps digitally leaves holes in the sound, which then have to be disguised.

In the first five reels, of music and effects only, there are only a few splices. For example, the overture, which plays over a black screen, is cut onto the rest of the reel. In reel three, the music track is continuous until the gunfight begins. At this point, there is a splice and the music continues with gunshots. The level of the music across the sections is quite even.

It is interesting that in reels four and five, changes of music are not accompanied by joins in the negative. In reel four, during the scene in the bar, a pianola is heard. However, the pianola begins after a moment’s silence. Similarly, in reel five the film's theme music begins again after the dance hall music has naturally concluded.

The limitations on sound mixing and the practicalities of dubbing explain why there are many more joins in the dialogue reels, 6-9. The beginning of reel 6 is a good example. There is a splice in the track after Warwick Ward's first piece of dialogue and before Katie answers him. This does coincide with a change of scene in the picture. However, there are several joins in her speech, even though she is the only one speaking. The levels vary noticeably between these sections.

How did we restore the sound?

The first step was to record the sound master digitally as 24 bit sound files.

The next stage was to remove pops and clicks and as much of the crackle as possible without damaging the audio. Some of this could be done with automated digital processes but a lot of the larger clicks and pops needed manual intervention to identify and remove them one by one. The heavy thumps produced by bloops needed a new technique to be developed in order to remove them without leaving gaps in the sound.

The next stage was to reduce the hiss and background noise inevitably introduced by the unsophisticated early recording equipment. On the later reels containing dialogue this could vary from shot to shot. In some speeches the low level of the voice recording and the high level of noise meant that forensic audio treatment was needed to recover the dialogue. Also at this stage sound levels were adjusted to give a more consistent overall volume. Some frequency equalisation was also applied to try to better match the sound of the speech elements between shots.

Finally improvements to the synchronisation of the sound effects in the first five reels and the lip sync of the dialogue in the last four reels were attempted. This meant re-cutting the soundtrack and shifting sections backwards and forwards to match the pictures. In some of the dialogue scenes sentences had to be chopped up syllable by syllable and reconstructed to sync the "ghost voice" with visible lip movements.

How did we restore the picture?

An examination of the safety fine grain was carried out to ascertain whether any digital repair work might be necessary. It was in good condition, so a duplicate negative was easily made, without the need of extensive grading. However, three sections were selected for digital repair.

Two of these were picture dissolves. The original negative was checked for these. The two sides of the dissolves no longer exist intact. This is unfortunate because the extant footage shows that fades had been made 'in-camera' before being incorporated into prints. At the time the fine grain was made, the dissolves had been combined and then cut badly into the bulk. The result was that the dissolves in the fine grain suffered from misalignment, 'flare', shifts in density and contrast and damage caused by editing cement. These sections were digitally scanned for restoration. Each side of the dissolve was created by duplicating the required number of frames then recreating the overlap. The other fault was a tear in Reel 2 that had been printed into the picture. This was digitally 'painted out'.

The corrected sections were then transferred back to film.

The end-product

This project combined traditional photochemical restoration techniques with digital restoration for intertitles, damage in the picture and above all to restore the sound.

We now have new master preservation material, and material from which further copies can be made if required. For screening, we have a new 35mm print and a digital copy on the very high quality High Definition format.

Andrea Kalas, Kieron Webb, Richard Yeoman-Clark
BFI National Archive Film, Video and Audio Lab

Thanks to Simon Hessel for his generous donation to the BFI, which made this restoration possible.

View All collections.