Return to site

Red Giant Software Unveils Film Fix Restoration Tools For Mac

broken image


My family, like many families in the days before video cameras, capturedsnippets of family history four minutes at a time on 8mm reels of film.Being a younger member of a large family, many of my memories of early familyhistory were formed while watching the old films on an old Bell & Howellprojector during the semi-annual home movie nights.

  1. Red Giant Software Unveils Film Fix Restoration Tools For Mac 2017
  2. Red Giant Software Unveils Film Fix Restoration Tools For Mac Free
  3. Red Giant Software Unveils Film Fix Restoration Tools For Mac Full
  4. Red Giant Software Unveils Film Fix Restoration Tools For Mac 2017

Around the time I was eleven or twelve, about junior high, I got interestedin making animated films, partly inspired by Terry Gilliam's crazy animationsscattered about various Monty Python episodes. Scraping together about twelvedollars for film and developing, I could make a movie with my brother or withfriends. These were largely unplanned productions, figured out after the filmhad been bought, the film loaded, and the 'talent' assembled. 'OK, what shouldwe do now?' was a common question. We didn't have a tripod, lighting was poor,and the results were, uh, 'organic' (like dung is organic), but were still alot of fun to make.

Jun 07, 2011  Film Fix A tool from Red Giant software. Target users are consumer to semi-pro. The product is discontinued. It appears Red Giant is working on a revamped version with no indication of when it will be re-introduced. It is a plug-in for After Effects v6 to CS3. I tried this tool in CS3.

  1. Depends how much money you have. There is Movie Looks for Vegas from Red Giant Software. VASST has collections of filter sets, many free, that might help a little. However most of the issues that make video look like video is inherent in the way you shoot and light.
  2. (San Francisco, CA, February 13, 2006) Red Giant Software, publisher of a line of professional desktop tools including Magic Bullet and Knoll Light Factory, today announced the availability of the Macintosh version of Film Fix. The plug-in supports Adobe After Effects 7 and is now cross-platform.
  3. Feb 12, 2006  February 11, 2006 - SiteTagger v2.1 released - It keeps getting better Nashville, TN - From Concentrate Software released SiteTagger v2.1. SiteTagger is a replacement for your browser's bookmarking functions. It features tag-based bookmark management, importing from most major browsers, and integration with several online bookmarking services.

In the summer of 2010, in my mother's basement I came across the boxcontaining all the family movies. It had been many years since I had seenthem. Years before a couple of my siblings had had the films transferred toVHS, but I had lost the tape. Film transfer technology has gotten much betterin the intervening years, so I decided then to have them recaptured.

This page is a summary of what I learned about it, and how I went aboutgetting a good quality restoration. I'm not saying this is the best way; itis just the way I did it and if you are considering doing something likethis yourself, maybe you can learn from my missteps. This summary is hardlycomprehensive.Hereis a good write up of many of the issues.

Added 2019: Note that this web page was written 2011, immediately afterfinishing up the film conversion project. I have not worked on it or thoughtmuch about it since. I am contacted, at least monthly, by someone asking if Iwill convert films for them. To head off more of that: sorry, the answer isno, I don't do that; I already have a full-time job and not enough free time.

Red giant software unveils film fix restoration tools for mac pc

Others ask for me to recommend a place to do the conversion. Because I haven'tlooked at all at this sector in 8 years, I really have no useful opinion.I'm sorry I can't be more help, but this article already contains all thethings I knew when I was somewhat up to date.

Interesting links on the same topic:

  • Fred Van de Putte's website (English)

Transfer Technology

Realtime transfer

Back in VHS days, one common method of transferring films to tape was quitesimple: the film would be projected on a screen, and a video camera would filmthe screen and capture the image in real time. Such transfers didn't requiremuch more than what any camera store already had on hand, and so many storesoffered such services. In fact, one can do this at home.

One problem is that the frame rate of films are different than the framerate of video. For instance, regular 8mm film is typically 16 fps, super 8mmfilms is typically 18 fps, and video is 30 fps (really 29.97 if you wantto nitpick). It is very possible to get strobing effects from the mismatchin frame rate. Another problem is that most projectors don't evenly lightthe frame: the center is brighter than the edges, producing a 'hot spot'effect. If a standard projector is used, the film gate (a piece of metalwith a rectangular hole in the center which the film is projected through)will block out all the image near the edges of each frame, as much as 20% ofthe image. If either the projector or the camera filming the image aren'tperpendicular to the projection surface,image distortionmay result.

There are slightly better units dedicated to the job that can remove someof these problems, but they still suffer from the problem of unsynchronizedframe rate conversion:

These days the image is captured to DVD instead of VHS, but there are the samefundamental flaws as before. Do not under any circumstance settle for thisoption, no matter how cheap.

Frame by Frame transfer

The next step up is 'frame by frame' transfer. In this setup, a modifiedprojector advances the film by one frame, and an image sensor mounted directlyin front of the lens of the projector captures one frame and the data istransferred to a computer. Then the film is advanced to the next frame,and the process repeats, typically doing ten or so frames per second. Bysynchronizing the display of each frame to the capture of each frame, thereis no flicker induced by having some frames captured while the film gate isclosed or only partially open.

In addition to the inherently better capture, software on the computer canperform image processing to remove defects like scratches, do color balancecorrection, and set light levels. Up-converting from the film's natural framerate to videos 30 fps can be done in an intelligent way to minimize thedistortions that may occur.

This approach is more expensive than real time transfer, but the quality isworth it. With good construction and careful operation, frame by frame capturecan rival the quality obtained by high end scanning equipment (describednext). However, nothing is guaranteed; some frame by frame equipment is notwell made, or the operator of the equipment might be careless or incompetent.

Here are a few links to people who make this type of equipment, some justfor personal use. Note that I have no experience with them and are justrepresentative of the type of technology discussed in this section. I willnote that Fred Van de Putte has produced some stunning results with his homemade equipment: How to purple restore for mac 2017.

  • Fred Van de Putte's telecine equipment, also see his videos on how he built his system, part 1 (hardware), and part 2 (capturing). View the other videos on his youtube channel to see how he has restored a few films using his capture system and avisynth for postprocessing.

High End Film Scanning

There are specially built machines for transferring film to digital format;these machines were first used by high end video production companies totransfer theatrical movies to digital format for broadcasters to use. Forexample, the Movie Channel doesn't actually broadcast old movies from film.The word you should use when doing searches on the subject is 'telecine.'

Like the 'frame by frame' capture just described, film scanners also captureimages a frame at a time, but in a more sophisticated way. Rather than takinga snapshot of each frame, like a camera, they scan the film a line at a timeor even a pixel at a time. At first blush this sounds like a step backwards,but rather than needing to capture one million pixels in one flash, filmscanners can afford to use a very high quality sensor because they need onlyone, or one row's worth. Such equipment is invariably part of a suite ofhardware dedicated to image processing, with trained operators to adjustlighting and correct color.

When discussing color correction, there are a few terms to learn. The bestquality, and most expensive, is 'scene by scene' correction. The operatorwill adjust lighting and color correction for each scene, and will ensureconsistency in the colors between different scenes. 'Best light' means thecolorist will make corrections for each scene, often on the fly, but withless attention to consistency. 'One light', the cheapest approach, means thecolorist will make a best compromise that seems to work across all scenes; inreality, if there is a dramatic shift in lighting or color, the colorist willlikely rewind and readjust at that point.

Another advantage of film scanners is they advance the film continuously,rather than using the typical sprocket advance and register scheme. This won'tmake a difference for some films, but many old films have bad splices andsprocket holes that are broken or 'chewed on' and won't feed reliably througha sprocket mechanism. Some old films can even shrink, leading to bad sprockethole spacing. The film scanner software has a means of registering each framewithout relying on sprocket holes to do the job.

Having been around for a long time, these high end film scanners are nowgetting cheap enough for use in capturing home movies. Some services use oldhigh end equipment from the 1990s; others use state of the art equipment, butyou pay for it. Either way, the quality is likely to be a cut above frameby frame capture equipment. To process 8mm film, these machines need customadapters, since they were designed for 16mm and 35mm work.

Rank Cintelscanners were the top end scanners, but theSpirit DataCine 4Know seems to be the (very expensive) top end. As far as I know, nobodydoes 8mm film transfers using Spirit-grade hardware.

Transfer Format

Chromecast for second monitor mac. No matter which capture scheme is used, eventually you need to decide in whatform do you want your digitized video to be returned to you.

Red Giant Software Unveils Film Fix Restoration Tools For Mac 2017

All such places offer to transfer you films directly to DVD. They can addtitles and even atmospheric music if you desire it. It is possible to grabthe video off of these DVDs for subsequent editing, but it is already heavilycompress MPEG2 video, and isn't convenient for editing. If you don't wantto mess with editing or pressing your own DVDs, then by all means, get yourcapture done directly to DVD.

Another option is to have the film transferred to aDVvideo tape; these tapes are digital format, and require some means oftransferring the bits from the tape into your PC for further editing. DV filesare reasonable for editing, as they offer frame level access with moderatecompression.

As I wanted maximum control over the end result, I opted to have my videocaptured and saved as a video files, one per film. I mailed a portable harddrive to the capture house and they returned my film as uncompressed AVIfiles. Uncompressed video results in huge files, but the advantage is thatthere are nocompression artifactsto worry about, and editing at an arbitrary frame level is no problem.

Red giant software unveils film fix restoration tools for mac windows 10

Others ask for me to recommend a place to do the conversion. Because I haven'tlooked at all at this sector in 8 years, I really have no useful opinion.I'm sorry I can't be more help, but this article already contains all thethings I knew when I was somewhat up to date.

Interesting links on the same topic:

  • Fred Van de Putte's website (English)

Transfer Technology

Realtime transfer

Back in VHS days, one common method of transferring films to tape was quitesimple: the film would be projected on a screen, and a video camera would filmthe screen and capture the image in real time. Such transfers didn't requiremuch more than what any camera store already had on hand, and so many storesoffered such services. In fact, one can do this at home.

One problem is that the frame rate of films are different than the framerate of video. For instance, regular 8mm film is typically 16 fps, super 8mmfilms is typically 18 fps, and video is 30 fps (really 29.97 if you wantto nitpick). It is very possible to get strobing effects from the mismatchin frame rate. Another problem is that most projectors don't evenly lightthe frame: the center is brighter than the edges, producing a 'hot spot'effect. If a standard projector is used, the film gate (a piece of metalwith a rectangular hole in the center which the film is projected through)will block out all the image near the edges of each frame, as much as 20% ofthe image. If either the projector or the camera filming the image aren'tperpendicular to the projection surface,image distortionmay result.

There are slightly better units dedicated to the job that can remove someof these problems, but they still suffer from the problem of unsynchronizedframe rate conversion:

These days the image is captured to DVD instead of VHS, but there are the samefundamental flaws as before. Do not under any circumstance settle for thisoption, no matter how cheap.

Frame by Frame transfer

The next step up is 'frame by frame' transfer. In this setup, a modifiedprojector advances the film by one frame, and an image sensor mounted directlyin front of the lens of the projector captures one frame and the data istransferred to a computer. Then the film is advanced to the next frame,and the process repeats, typically doing ten or so frames per second. Bysynchronizing the display of each frame to the capture of each frame, thereis no flicker induced by having some frames captured while the film gate isclosed or only partially open.

In addition to the inherently better capture, software on the computer canperform image processing to remove defects like scratches, do color balancecorrection, and set light levels. Up-converting from the film's natural framerate to videos 30 fps can be done in an intelligent way to minimize thedistortions that may occur.

This approach is more expensive than real time transfer, but the quality isworth it. With good construction and careful operation, frame by frame capturecan rival the quality obtained by high end scanning equipment (describednext). However, nothing is guaranteed; some frame by frame equipment is notwell made, or the operator of the equipment might be careless or incompetent.

Here are a few links to people who make this type of equipment, some justfor personal use. Note that I have no experience with them and are justrepresentative of the type of technology discussed in this section. I willnote that Fred Van de Putte has produced some stunning results with his homemade equipment: How to purple restore for mac 2017.

  • Fred Van de Putte's telecine equipment, also see his videos on how he built his system, part 1 (hardware), and part 2 (capturing). View the other videos on his youtube channel to see how he has restored a few films using his capture system and avisynth for postprocessing.

High End Film Scanning

There are specially built machines for transferring film to digital format;these machines were first used by high end video production companies totransfer theatrical movies to digital format for broadcasters to use. Forexample, the Movie Channel doesn't actually broadcast old movies from film.The word you should use when doing searches on the subject is 'telecine.'

Like the 'frame by frame' capture just described, film scanners also captureimages a frame at a time, but in a more sophisticated way. Rather than takinga snapshot of each frame, like a camera, they scan the film a line at a timeor even a pixel at a time. At first blush this sounds like a step backwards,but rather than needing to capture one million pixels in one flash, filmscanners can afford to use a very high quality sensor because they need onlyone, or one row's worth. Such equipment is invariably part of a suite ofhardware dedicated to image processing, with trained operators to adjustlighting and correct color.

When discussing color correction, there are a few terms to learn. The bestquality, and most expensive, is 'scene by scene' correction. The operatorwill adjust lighting and color correction for each scene, and will ensureconsistency in the colors between different scenes. 'Best light' means thecolorist will make corrections for each scene, often on the fly, but withless attention to consistency. 'One light', the cheapest approach, means thecolorist will make a best compromise that seems to work across all scenes; inreality, if there is a dramatic shift in lighting or color, the colorist willlikely rewind and readjust at that point.

Another advantage of film scanners is they advance the film continuously,rather than using the typical sprocket advance and register scheme. This won'tmake a difference for some films, but many old films have bad splices andsprocket holes that are broken or 'chewed on' and won't feed reliably througha sprocket mechanism. Some old films can even shrink, leading to bad sprockethole spacing. The film scanner software has a means of registering each framewithout relying on sprocket holes to do the job.

Having been around for a long time, these high end film scanners are nowgetting cheap enough for use in capturing home movies. Some services use oldhigh end equipment from the 1990s; others use state of the art equipment, butyou pay for it. Either way, the quality is likely to be a cut above frameby frame capture equipment. To process 8mm film, these machines need customadapters, since they were designed for 16mm and 35mm work.

Rank Cintelscanners were the top end scanners, but theSpirit DataCine 4Know seems to be the (very expensive) top end. As far as I know, nobodydoes 8mm film transfers using Spirit-grade hardware.

Transfer Format

Chromecast for second monitor mac. No matter which capture scheme is used, eventually you need to decide in whatform do you want your digitized video to be returned to you.

Red Giant Software Unveils Film Fix Restoration Tools For Mac 2017

All such places offer to transfer you films directly to DVD. They can addtitles and even atmospheric music if you desire it. It is possible to grabthe video off of these DVDs for subsequent editing, but it is already heavilycompress MPEG2 video, and isn't convenient for editing. If you don't wantto mess with editing or pressing your own DVDs, then by all means, get yourcapture done directly to DVD.

Another option is to have the film transferred to aDVvideo tape; these tapes are digital format, and require some means oftransferring the bits from the tape into your PC for further editing. DV filesare reasonable for editing, as they offer frame level access with moderatecompression.

As I wanted maximum control over the end result, I opted to have my videocaptured and saved as a video files, one per film. I mailed a portable harddrive to the capture house and they returned my film as uncompressed AVIfiles. Uncompressed video results in huge files, but the advantage is thatthere are nocompression artifactsto worry about, and editing at an arbitrary frame level is no problem.

As a comparison, a standard definitionMPEG 2video found on a DVD is typically encoded at 3.5 to 8.0 Mbit/s;DV is 25 Mbit/s;uncompressed video is around 240 Mbit/s.

If you opt for a file-based transfer on a hard disk, there are more questionsto be answered. What resolution should the film be captured? What videocontainer format should be used? What type of code should be used to put thevideo in the file? Should the film be captured with pulldown? Interlaced orprogressive format? Some of these subjects (such as choice of codec) requiretheir own web page, but here is a very, very brief summary of some of theseconsiderations.

There is a temptation to capture the film at the highest offered resolution.It does cost more, though, and processing the results is a lot slower thancapturing at standard definition resolution (720x480). Keep in mind that 8mmframes are quite small (less than 8mm!) and don't inherently contain more thana certain amount of information. Getting 1920x1080 capture of such film willjust give you very clear pictures of the film grain. Personally, going higherthan 1024x768 is overkill for 8mm.

As was mentioned, the native film rate is 16 fps or 18 fps for 8mm film, 24fps for 16 mm film, and 30 fps for video. One way to convert the native filmframe rate to the video frame rate is to duplicate some frames in the videostream. That is OK for playback, but it makes processing the video difficult.Ask the transfer house to give you one frame of video for each frame on thefilm. You can do frame rate conversion after you have done all of your filmediting and enhancement.

TV has historically presented video ininterlacedformat, presented at 29.97 fps (in North America). However, film itselfis inherentlyprogressive.Deinterlacingsoftware can be used to convert interlaced back to progressive format.If you have a choice, stick with progressive scan, as it makes subsequentimage processing tasks easier.

If you use an Apple Mac computer for editing, you probably want to usequicktime MOV format as the file container. Using theProRes HQcodec will give you the highest quality transfer, as it doesn't compress thedata much -- only subsampling the chroma by a factor of two.

If you use a PC, you'll probably want to use an AVI file container. Fora codec, you should avoid anything that applies a lot of compression.Highly compressed files are relatively small, but leads to quality loss anddifficulty in editing. You want to save compression to the very end of yourchain of editing tools.MJPG(motion jpeg) is a fine format, as it doesn't have any form of interframecompression, allowing for easy access to arbitrary frame boundaries. Thereare some lossless compressors, likehuffyuv andlagarith.Apple ProRes video is also known as v210 by itsfourcccode, and there are PC codecs for reading this uncompressed format.

No matter which format you pick, you should first ask for a small (like 10frame) sample of film encoded in that format from your transfer house so youcan verify that you are able to open it and recover the images. It would beterrible to spend $600 capturing a bunch of movies only to find that you can'tread them.

Cleaning

Old films can be dusty, or have an accumulation of grime from a lifetime ofhandling. Old films often have suffered scratches; which side of the film isscratched makes a difference.

Cheaper transfers simply load your old film and go. The next step up is tohave the film 'dusted' by mechanically passing it between brushes to try andknock off any dust or dirt. Better quality cleaning has the film pass betweentwo sponges wetted with a cleaning solution; the solution is more active thanwater, but not so much that it will damage the film.

There is also something called a 'wet gate' (or 'wetgate') transfer. Toexplain it, we need some background information. The actual film emulsion ison side of the plastic backing. If the scratch is on the emulsion side, itoften shows up as a blue streak in the film; the film is truly damaged and thiscan't be fixed during the capture (although it can be fixed inpost-processing). If the scratch is on the backing side, it shows up as a darkscratch when projected. The wetgate process means the film is coated with aliquid which has an index of refraction matching that of the plastic; thisliquid fills in the valley formed by a scratch, and the film emulsion can thenbe captured properly. It might sounds like hokum, but it really works. Belowis a youtube video showing the effect on a particularly badly scratched film.

Cost

Red Giant Software Unveils Film Fix Restoration Tools For Mac Free

In general, quality costs money. Real time transfer will be the leastexpensive, frame-by-frame capture more expensive, and film scanning will bemore expensive yet. There is a large variation in pricing, and the cost perfoot drops if you have a large order. However, to give you some estimate, realtime capture can be $4-$8 for a 50 foot reel of 8mm film; older scanningequipment can run $15-$25 for a 50 foot reel. I've seen quotes as high as $40per 50 foot reel; as I didn't opt for that service, I don't know if it deliverssuperior results or is simply more expensive.

In addition to the cost of the transfer, usually based on the number of feet offilm, there are often other charges. There can be preparation charge to repairbad splices or bad leaders. The higher end transfer equipment is veryexpensive to operate and they don't want to waste a lot of time messing aroundwith 50 ft. film reels. Instead, all of your reels get joined together on asingle 400 ft. reel which can then be efficiently processed on the expensivescanning equipment. There can be a charge for cleaning the film, and a chargefor transferring the result onto DVD, tape, or hard drive. Capturing soundfilms is more expensive than silent films. Some places charge a fixed overheadper job, perhaps to scare away the small orders.

I recently had about 35 reels of silent 8mm film captured at standard defresolution, saved to .avi files. In fact, I had it done twice. The first timewas at a local video transfer place using frame by frame capture. In total, itwas about $300. The nominal price was $7.50 per 50 foot roll, but there weresome other fees that raised the total.

After struggling to improve the quality in post-processing, I bit the bulletand had it all redone atCinepost.Due to the volume of films processed, the cost was a very reasonable $15 per 50ft. reel, plus some other reasonable overhead expenses. In total it was about$700, but, as you'll see, it was well worth it. Cinepost uses Cintel filmscanners with some custom hardware that allows them to handle 8mm film.The turn around time was about a week.

Image Comparison

I don't have a lot of experience, in fact, just two data points. But I wantedto share what I got back from the mom & pop video service I used(frame-by-frame capture) vs. when I went to a professional telecine service(Cinepost). Or to put it another way, what does 15¢/foot processinggive you vs. 30¢/foot. These are the corresponding frames of the samefilm from the two captures. Note that alignment isn't exactly the same. Notetoo that I had asked both to do a wide capture including the sprocket holes;normally they crop the video so the sprocket hole doesn't show up. I prefer tothe cropping myself.

Capture from a cheap frame-by-frame service:

Capture using a Rank Cintel film scanner, without post-processing,done at Cinepost:

It should be stated that normally Cinepost does post processing on the imageafter capture, but I requested that it not be done, as I wanted to do it allmyself. Normally, Cinepost would return better looking results than this.

The most obvious difference vs. the cheap capture is that, even in this rawstate, the colors are much more natural. Look too at the bright reflection offthe stroller handle on the right edge of the frame: the cheaper capture bleedsbrightness into the area next to the handle. The Cinepost capture has muchmore uniform lighting. Look at the yellow and green chairs at the top of theframe: the image is much crisper; this is evident too in the wires of the chainlink fence.

There is something else that is instructive in this capture. Look at thecolor of the light coming through the sprocket holes on the left side ofthe frame. In the cheap capture, it is white light; in the professionallydone one, the illuminating light color as well as intensity is chosen tocompensate for any color shifts of the aging film.

Image Enhancement

Many people receive captures like the first one shown on a DVD; that is thatand they just live with the poor quality. Much can be done to improve thequality in post processing, but it is always best to start with the bestcapture you can.

vReveal

First I tried a piece of software namedvReveal,made by a company named MotionDSP. MotionDSP makes somevery expensive softwarefor forensic video restoration. vReveal is a consumer version of it, but withmany fewer knobs, at a very reasonable price point (US$39 as of January, 2011).There are only a few knobs for controlling what algorithms are applied to theimage, such as motion stabilization, contrast adjustment, fill lighting,sharpening, etc. What it does do, though, it generally does well. They havea free version you can download and use, but it adds a small vReveal watermarkon the processed video.

Here is a same Cinepost-captured frame as above after vReveal has enhanced it:

One of the things vReveal doesn't do is allow changing the settings during amovie. This can be worked around by chopping a film into one small video perscene, then tweaking the settings for each scene, and pasting the results backtogether, but that would be more than tedious. Also, some of the knobs are toocoarse; for example, the 'clean' function is either 'off,' 'on,' or 'high.'Whatever its shortcomings, though, it couldn't be easier to use.

Avisynth

I probably could have used vReveal on my films and been done quickly.However, I like to tinker, so I took a different approach, one populatedwith pitfalls and fraught with frustration.

Avisynthis an open source software product that runs only under Microsoft Windows.It has no commercial support, development advances in fits and starts,and despitea healthy user community,you will need to figure out a lot of things through trial and error.

Normally if a file is named something like foo.avi and you open it underwindows, it knows how to decode the AVI container format. Inside arestreams of data for video and audio, each of which must be decoded usingthe codec specific to that stream type. The data is processed by the codecand fed back to the application as a raw stream of pixels or audio samples,which are then displayed or what have you.

Avisynth is not an application as much as it is a driver. Conventionally,avisynth files have a suffix of '.AVS'. Existing video for windows-awareapplications can open .avs files and display the video. The strange thingis that .avs files don't contain any video! Instead, they are scripts thatdescribe how to generate video. When an application opens the file and asksfor frame 0, the avisynth driver reads the script, figures out how to generateframe zero, then feeds the decoded video to the application, which thinks thevideo had simply been read from a file.

There are plenty of tutorials around that describe how to use avisynth,but here is a small example to give a flavor of what it looks like.

This (untested) script opens the file named 'myvideo.avi', converts it to a32-bit RGB pixel format, and selects a 400x300 square of pixels from each frameoffset from the top by 100 pixels and the left by 100 pixels. Next it extractstwo scenes, the first from frame 622 to 894, and the second from frame 100 to434. Notice that the order of the scenes has been reversed from the original.Avisynth is a non-linear editor, meaning it isn't constrained to process framesin the order they appear in the file. These two scenes are then resized to be320x240 pixels, its brightness is boosted by 30 units, color saturation isenhanced by 20%, and the contrast is boosted by 10%. The result is returnedfrom the script. There are many ways to write this script, as the syntax issomewhat flexible.

One of the strengths of avisynth is that it has an open architecture wherenew filters (processing algorithms) can be created by anybody and pluggedinto the avisynth framework as seamlessly as if they had been part of theoriginal program. One of the weaknesses of avisynth is there are myriadfilters, some good, some bad, some compatible, some broken, and it takes alot of time to figure out what plays well together. It is also dauntingthat in effect it is a tool with 10,000 knobs, a potentially paralyzingnumber of options.

Avisynth can do other things than just video processing. It can overlayvideos, merge and process audio, change frame rates, change video formats,do titling, display analysis data, and more.

The thing that helped me tremendously was a script written by a user ofavisynth,Fred Van de Putte.Fred not only built his own telecine equipment, he has invested hundreds ofhours tweaking scripts to achieve the best possible results. Even so, eachfilm requires its own knob tuning, so the script doesn't make things nearly aseasy as vReveal does. The script and extensive dialog about using it can befound inone threadof a forum where things like this are discussed. The very first post in thethread is kept up to date with the latest version of Fred's script, and therearelinks to some of his results,which are well worth viewing to see just how high quality 8mm film can beafter restoration.

Here is the same frame as before, this time restored using Fred's script:

As compared to the vReveal results, note that the colors are more lifelikeand the image is sharper. There are other benefits as well, but this oneframe can't show them all.

Restoration Details

Again,this threadis the best place to read up on Fred's enhancement script, but this sectionwill give an overview of what it does. The steps are performed in theorder listed here:

Stabilize

All of my movies were filmed with a handheld camera, which resulted insubstantial amounts of shake and uneven panning. The script uses somesophisticated algorithms to detect what part of the scene is background(that is, unchanging) image and what parts are moving objects. This canbe tricky considering that when the camera pans, in some naive senseeverything is moving relative to where it was in the previous frame.With this information, the script looks at a sequence of frames and adjustseach one's position to smooth out any short-term jitter or shake and tosmooth the trajectory of intentional pans.

Crop

The film needs to be cropped to just the essential portions at this point.There are two reasons why cropping is needed. First, the original capture mayhave caught some of the border around the frame of the film and maybe somesprocket holes. Second, when the motion stabilization step, above, shifts eachframe to smooth out the image, one or two edges of the image must be filled inwith something to make up for the part of the image which was shifted into theframe (an equal amount on the opposite side was shifted out of the frame).

Clean

Even if the film was cleaned before capture, various specks of dirt mightstill adhere to the frame, or bits of emulsion might have flaked off yearsbefore, resulting in splotches that appear for one frame and then disappearagain. This step does something very sophisticated to fix these problems.For each frame, it compares the corresponding parts of the frame to theframes immediately before and after it. If it finds something which lookslike dirt (of a certain size, of a too bright or too dark color) whichappears for a single frame, it can repair cover up that spot by borrowingthe image from the corresponding place from the adjacent frames.

This is a little bit tricky as objects move from frame to frame. If awhite ball moves swiftly across a scene, it takes some intelligence totell that it is really one moving object and not a series of white specksthat happen in a different place in successive frames. Sometimes theheuristics that protect against things like this don't work and small,rapidly moving objects seem to disappear! Thankfully it is rare, and canusually be fixed up by adjusting the strength of the cleaning algorithm.

Here is a short clip showing how cleaning and denoising (next) can remove theframe gunk. Notice the blotch on the upper right part of the statue pedestal isremoved, and the smudges near the middle top part of the frame are gone.The right side is notably sharper due to the way the image processing scriptwas written, but isn't an inherent part of dirt removal.

(click on the image to see a short video clip)

(this clip of the 1939 World's Fair is from the Prelinger Archives)

Denoise

Denoising is different than cleaning, but it operates on some of the sameprinciples that the cleaner does. Film grain contributes a certain amountof noise to the image. By comparing a series of adjacent frames, thealgorithm identifies the corresponding point of an object in each of theframes. It can then do some intelligent averaging to discern what portionof the signal is due to the inherent attributes of the object being filmedand what is just contributed by film grain, so that film grain can be reduced.Removing too much grain causes the film to lose some of its character, andcan lose fine detail, resulting in a plastic look.

Sharpen

Some of the previous steps tend to soften the image somewhat; this stepapplies a filter to restore some of the sharpness, paying special attentionto the edges of objects where sharp transitions are the most noticeable.Like anything, sharpening can be overdone. Sharpening amplifies noise,and sharp edges can produce ghost images near the edge, called 'ringing.'

Frame rate conversion

Here is the point where the native frame rate of the film, be it 16 fps,18 fps, or something else, gets converted to 30 fps. Playing a 16 fps sourceat 30 fps would make everything happen too quickly. One approach is toduplicate frames until the right speed is achieved. For instance, say onesecond of film is represented by this sequence of 16 frames, named 'A' to 'P':

can be turned into these thirty frames via duplication such that playbackat 30fps takes one second, and the speed of the movie is unchanged:

Each frame was duplicated, except for frame 'H' and 'P', otherwise we wouldhave ended up with 32 frames. Because some frames are duplicated and othersaren't, the result will be somewhat uneven speed, a slight stuttering effect.Even if every frame is is duplicated the same number of times, the resultingvideo is more choppy than what video normally looks like, and is especiallynoticeable during panning.

Another approach is to go ahead and duplicate all 16 frames, resulting in 32,and then playing it at 30 fps. The motion will be about 6% too slow, butwill be more smooth.

Somewhat more sophisticated, it is possible to blend adjacent frames, weightingthe contribution of each based on how close it is in time to the frame whichis being synthesized. For instance, if we were simply trying to double theframe rate, we could average together frames A and B producing frame A', andthen play them back in A,A',B order.

Avisynth offers another option still, the most sophisticated of them all.Often it gives stunning results, but it can also produce noticeable artifactsand occasionally comically bad results. Avisynth can actually analyze eachframe and discern objects from other objects based on how they move inrelation to the background and each other, and how they occlude other objects.The tool can create a synthetic image part way between two frames in aphysically plausible manner. Imagine someone standing still but raisingtheir arm. The blending approach would produce an intermediate frameshowing a ghost of the arm in both the before and after positions. Frameinterpolation would try to create an image with the arm in a half waybetween position. Keep in mind that the tool is just shifting around pixelsand has no a priori understanding of objects, nor any understanding thatthe objects are originate from 3D objects projected onto a 2D image.Sometimes the heuristics can't figure out what to do, and it falls backon just doing a blending operation. Objects sometimes have a 'halo' ofblurred image around moving objects where it has synthesized some backgroundimage around a moving object.

Red Giant Software Unveils Film Fix Restoration Tools For Mac Full

Here is an extreme example: three girls are playing ring around the rosie, andthe film has been slowed down by a factor of 100. That is, for each pair oforiginal frames, 99 intermediate frames are synthesized and placed between thetwo real frames. The left portion is generated via frame duplication: eachframe is simply repeated 100 times in a row. In the middle is frame blending:notice how the left edge of the white dress fades in to advance. On the rightis interpolation using sophisticated motion analysis: the dress moves toadvance; on the other hand, the motion of the girl's legs is less convincing.

The same thing is presented below, but running at real time. This conversionjust changes the video from 15 fps to 30 fps. The stuttering effect isn't allthat strong in this clip, but it is visible when viewed closely. The frameblending in the middle produces a slightly blurry image for things in motion.The motion interpolation on the right looks the best of the three.

Color and levels adjustment

Finally, here is the place where the brightness, contrast, and gamma canbe twiddled, and the color corrected. The script has options for attemptingto do this automatically, or one can just leave it unchanged and useanother tool for adjusting these details.

Finished Product

Here is about a minute of the finished product. This is from Thanksgivingdinner, November 1968.

Other Links

Red Giant Software Unveils Film Fix Restoration Tools For Mac 2017

Here are a few other articles I've found since writing this; the list isn'texhaustive:

  • An general intro, like this write up
  • PC Magazine did a quality comparison between different transfer technologies

Feedback

If you want to contact me for whatever reason, try me atjim@thebattles.net.

Back to Jim's Junk main page.

Last updated: March 2, 2019





broken image