Vintage nitrate film presents dilemmas for the digital restoration artist

Given the time, digital tools and artistic/restorative abilities, this series of frames from 35mm nitrate could be restored to near perfect condition.  But 'could be' maybe is not the question.  The question  is rather - should certain segments be restored at all?

As we look back over 100-years at media that can never again be the way it would have been the day the film was first developed - wouldn't leaving specific severely degraded frames 'as-is' as part of a larger restored film be an reminder to the viewer of the reality that the media is ephemeral?

These are the questions I am asking myself when restoring this feature-length studio film from the silent era.

Comments welcome.

 

Expectations: Restored Nitrate Film

Today's silent-era film restoration starts with high-definition digital scan of a nitrate print.  In the case of "The Trail to Yesterday" (1918), the only surviving print at the eyeFILM Museum in Amsterdam was scanned at 2K resolution in 10-bit DPX digital format.  

Nitrate based film stock of the silent era was the first transparent plasticized medium with a photosensitive emulsion that first became commercially available by Eastman in 1889 and the was primary film stock for motion picture until 1951 when the movie industry transitioned to cellulose acetate film, or safety film.

My restoration work flow is a 2-pass process. The first pass uses VivaPRO software by Algosoft, that uses mathematically-based artificial intelligence models to precisely measure, analyze and correct levels of grain, noise, flicker, color representation, luminescence, chromatic properties, artifacts and other imperfections in every frame.

The second-pass is a manual process to remove any remaining artifacts or imperfections, using additional software tools.

A side-by-side comparison is presented below.  The film frame on the left is the starting point - the high-definition scan of the original nitrate film stock. The frame on the right is a result of both the first and second-pass processes.  

CLICK ON ANY IMAGE TO ENLARGE

The attention or emphasis in the 'cleaned' frame is not only artifact removal, but the resulting texture of the AI-driven process to reduce noise and provide a more lucid image.  

Film restorers must ask the question as to whether to add realistic film grain back into the images after cleaning. For this film, that is still an open question.

The image below is an extreme close-up the actor Anna Q Nilsson.  Here is where the noise removal and other processing from the VivaPRO software is effective in providing a restored look to the image.  

Thus the result of the AI-driven analysis and corrective action, is a both a pleasing and cinematic look for the restored film. 

 

On The Topic of 'Intent' in Restoration

My main objective in the restoration of 'The Trail To Yesterday' is to remove to the extent possible any damage or decay to the film that would tend to distract the viewer from the characters or visual intent of the movie.

 

CLICK ON ANY IMAGE TO ENLARGE

Above: (screen-shot): Each rectangle is an area of the film that required repair using various digital restoration tools.  A single frame in this film is 1/16th of a second.

A Restored Frame

This posting is meant to show how almost every frame is crafted as part of the restoration process.  The initial first-pass uses an artificial intelligence based 'automatic' process using advanced software.  The frame below is the result of that initial first pass.  Looking closely you can see residual artifacts that need to be removed.  An artifact on the woman's face is an example.

 

 CLICK ON ANY IMAGE TO ENLARGE

The white outlined boxes in the frame below show areas of the frame that were restored using manual tools found in the studio version of DaVinci Resolve software.  There are various methods to repair defects in digitally scammed footage.

 

 

 The resulting frame (below) is the final result.  At first it doesn't seem like much change, but one of my goals is to keep the viewer's attention on the character and scene and minimize distraction from momentary blemishes on this aging film.

 

Challenging Scene Restoration

 

This is a restoration example of a very short scene (16-frames, approximately 1-second of film) from the continuing restoration of The Trail to Yesterday being conducted by Bart Santello and Psychotropic Films LLC. 

As you can see from 'as-scanned' not restored footage (left frame), this sequence of frames was damaged and degraded resulting in 'jittering' of the frame as it indexed through the scanner.  And as you watch the sequence it looks like there are some missing frames (jump-ahead).

In this case the each frame of the DPX sequence was first manually realigned in DaVinci Resolve prior to use of automatic processing software - VIVA by Algosoft-Tech, that processes each DPX frame using artificial intelligence and neural network technology to clean and restore.  Several passes with adjusted setting were employed using VIVA to ensure a deep clean (dust busting and Luminance correction).

Any remaining artifacts were manually removed in DaVinci Resolve along with final stabilization.  Again, the result was excellent compared to the original nitrate footage scan. 

Restoration of Split-Frame Scene

This is a restoration example of a very short scene from the continuing restoration of The Trail to Yesterday being conducted by Bart Santello and Psychotropic Films LLC. 

Due to deterioration of sprocket holes in this 100-year old Nitrate print, mis-alignment during the digital scan resulted in a split frame sequence.

Before the restoration could begin, each frame of the scene was digitally realigned in Davinci Resolve (video editor). 

The first pass of the restoration used software by Algosoft-Tech (VIVA), that processes each DPX frame using artificial intelligence and neural network technology. Then the result was put back into Davinci Resolve Studio for manual restoration to clean up any remaining artifacts. 

I started with this very short scene in order to develop a process for other scenes in the film with a similar split-frame issue.

The following slides are part of the historical research I'm doing for background into the production of "A Trail to Yesterday".  I clipped news articles on the film from trade and entertainment magazines at the time.  There were even 'inside the industry' and 'rumor' columns where just a sentence or two could reveal great insight on where, why and hows of the production.   And leads to new information which then becomes stories in their own right.

Another Mystery of the Production

While researching digitized movie industry trade journals from the time of  The Trail To Yesterday, I came across this mention in Motion Picture News touting a scene that is not in the film copy I have.   There is a scene in the film where the heroine Actress Anna Q. Nilsson sees a snake while on her horse, leading the horse to a wild gallop as she faints. Then rescued by the lead Bert Lytell.   

When I first saw this scene after reviewing the digitized film from the archive in Amsterdam, my first thought was that the 'quick-sand' scene (as described in Charles Alden Seltzer's book version - 1913)  was replaced with the snake scene in the movie.  This makes sense since both in the book and the movie this marks a turning-point where the heroine begins to trust the cowboy Dakota (Bert Lytell).

So why did this trade journal 'rumor' not come to fruition in the film?  My speculation is that this film shot in March and screened in May of 1918.  The studios at the time cranked out many films in a year.  From my research, the star actor, Bert Lytell, was in four pictures between March and August.

Therefore, the extra time and production cost to do a quick-sand scene may have lead to it being scrapped and replaced with simply calling-in a snake wrangler, placing the snake near the horse and setting up the photography.

In addition, I believe it would have not been edited out of this foreign distributed version of the film; since a quick-sand scene would be considered a high-value scene.

Hopefully further research will lead to a conclusive answer to this mini-mystery.

 Archived Motion Picture News clipping on this topic.

 

Film Restoration.....Eye of the beholder

This is my 'workspace' in DaVinci Resolve's editing environment for the next step in the process of restoring of "The Trail to Yesterday". [A Metro Pictures Corporation, 35mm film release in 1918].

As I move frame-to-frame, I further remove manually any remaining artifacts from the film's decay; or, scratches and other handling defects the 'first-pass' automatic AI-based software method didn't catch. 

I have to think about what level of completeness in the restoration am I willing to undertake to achieve the quality I expect from myself.  Every frame is 1/16th of a second so I know that the mind's-eye won't be able to recognize a certain level of detail. 

I haven't come to terms yet that I may need to examine the time involved crafting each frame with the realization that such a level to detail involves time.  Is that last 10-15% of 'quality' in the digital film repair, visually relevant to the viewer's perception?  

The opportunity to bring something back to consciousness after over 100-years of being stored in an archive and essentially forgotten, brings an overwhelming sense of responsibility to do right with the media restoration and production, before being returned to society as a restored work.

"Trail to Yesterday" restoration showcased in presentation on Artifical Intelligence

Psychotropic Films is employing the latest software technology that utilizes artificial intelligence (AI) to assist in the analysis and repair (restoration) of  “The Trail to Yesterday” (1918).  The following presentation by Algosoft-Tech (below) discusses the latest advances in AI and how AI is utilized in software to locate and correct defects in film. 

You can see in this presentation at 35:45-min, a featured a scene from The Trial to Yesterday  demonstrating the effect of the Algo-Soft software to assist in the restoration of film scanned to digital from 104 year old 35mm nitrate film stock.



How Recent 'AI' Breakthroughs Are Transforming Moving Image Restoration from Psychotropic Films on Vimeo.