FILM9 - Restoration software for old films (8mm, Super8, 16mm) and video

[sholee]

I can't count anymore how many times I have removed the "Film9v2" folder in %appdata% and the "Film9" folder in "program files (x86)" since I am trying to install this program

Yes, I did a search in the registry, but looking only for entries related to "film9" and there weren't many things left after uninstallation.
Are you thinking of specific registry folders I should look into? Please don't expect me to use programs like "CCleaner" as they can make things often worse This is something I learned very early during my 20 years+ in the IT-Business.

Yes, there are some programs installed as I do professional mixing and mastering of audio recordings for other studios. So there is everything that is state-of-the-art in studios around the world and is needed to collaborate with other studios. Cubase, Sound-Forge, Melodyne, Har-Bal and so on including a bunch of virtual instruments and plug-ins.

Videoediting and -restauration is a long time hobby, so you might find DaVinci Resolve, Hitfilm Express, Vegas Pro, and some little tools like Avidemux, Handbrake, VirtualDub 2 etc

Sometimes I need to edit pictures manually (captured images of super8 home movies in *.jpg), so you might find programs like Gimp, Topaz Denoise, Topaz GigaPixel, Topaz Studio, IrfanView etc

And, yes, there are a few games installed, because sometimes I need something to flee the reality

But Film 9 was the first program where I encountered problems and the first program where I allowed the installation of a "codec pack" (!) I remember 10 years ago, when the K-Lite Codec Pack made a lot of nonsense to my system and I had to reinstall Windows to make things work again. So I was not very overwhelmed when I had to install this codec pack. But I did some research and found out that it is safe to install and so far it did no harm to my other audio/video programs

Codecs I installed prior to Film9 were:

- Lagarith
- x264vfw (which I uninstalled before installation of Film9)
- Grassvalley Lossles
- Huffyuv 2.1.1
- ffmpeg
- MagicYUV (the old free version)
- UtVideo

At this time I'm setting up another old laptop with a clean installation of Windows 10 v1809 to see if this works with Film9 ...ThumbsUp ^^

[jammin]

I just installed version 2.14 - first time i've installed any version of film9 before.

However it won't start. I can see the process runs for a few seconds, but I see no UI.

I wasn't prompted to install any additional applications or codecs during the installation. I am running Windows 10 with all updates.

I have uninstalled / reinstalled. I have tried manually installing avisynth.

In the windows application log I see 2 errors:

Application: Film9.exe
Framework Version: v4.0.30319
Description: The process was terminated due to an unhandled exception.
Exception Info: System.IO.DirectoryNotFoundException
   at System.IO.__Error.WinIOError(Int32, System.String)
   at System.IO.FileStream.Init(System.String, System.IO.FileMode, System.IO.FileAccess, Int32, Boolean, System.IO.FileShare, Int32, System.IO.FileOptions, SECURITY_ATTRIBUTES, System.String, Boolean, Boolean, Boolean)
   at System.IO.FileStream..ctor(System.String, System.IO.FileMode, System.IO.FileAccess, System.IO.FileShare, Int32, System.IO.FileOptions, System.String, Boolean, Boolean, Boolean)
   at System.IO.File.InternalWriteAllBytes(System.String, Byte[], Boolean)
   at System.IO.File.WriteAllBytes(System.String, Byte[])
   at l.a()
   at h.a()

Faulting application name: Film9.exe, version: 2.0.0.0, time stamp: 0x5c2610ea
Faulting module name: KERNELBASE.dll, version: 10.0.17134.441, time stamp: 0x3da51fd0
Exception code: 0xe0434352
Fault offset: 0x00111812
Faulting process id: 0x4754
Faulting application start time: 0x01d4a337f2f6b24a
Faulting application path: C:\Program Files (x86)\Film9\Film9.exe
Faulting module path: C:\WINDOWS\System32\KERNELBASE.dll
Report Id: 3cb03abc-1fdd-4b1f-97a2-7e28d6b3ff84
Faulting package full name: 
Faulting package-relative application ID:

Any ideas?

Thanks!

[Gelinox]

Hello,

A new version is available:

Version 2.15 of 31/12/2018
Improved compatibility with Windows10
Fixed a bug on the detection of an existing directory

This version fixes the bug that prevents opening the application when using Windows10.
Still on Windows10, it no longer requires the installation of a dedicated Framework.

On the other hand, an installation of Framework 4.7 will be proposed (if not already installed) on the computers using an older OS.

This update addresses the issues mentioned, but unfortunately not yet those of Sholee.
We are still looking for the cause.

Gilles et Roland

[jammin]

A new version is available:
...
This update addresses the issues mentioned, but unfortunately not yet those of Sholee.

Merci Beucoup! Just to confirm this has fixed my problem. Now I just need to work out how to use it

[mcconnor]

If you use the film restoration script that VideoFred developed, and which I then adapted and sped up, you can sometimes remove certain scratches by using a really high setting for RemoveDirtMC.

You'll find all these film restoration scripts, including my versions, in these two threads:

The power of Avisynth restoring old 8mm films

Capturing and restoring old 8mm films

You might first want to try this script that StainlessS knocked together. The OP, who was also trying to remove film scratches, reported success. I have not tried it.

Removing thin vertical lines from Video

The script is given in post #7.

Hello johnmeyer,

Just remembered this clue you gave me and realized I probably did place my recent post concerning advice in this matter in the wrong (rather old and therefore probably "off-the-screen") thread. I ran into a wall there and perhaps you or some other experienced video-restorer could give me insight as how to proceed (if so...). Thanks for any advice in my request here (I don't want to cross post): https://forum.videohelp.com/threads/378183-8mm-restoration-script-question/page2#post2539232

[mcconnor]

is it just me or does exporting a project in FFV1 format actually produce uncompressed video? Just tried that today and got a 430 Gb avi where I expected an about 130 Gb avi. Double checked in the project settings: avi-FFV1 is selected...

(Film9 2.15)

[mcconnor]

Hello,


This version fixes the bug that prevents opening the application when using Windows10.
Still on Windows10, it no longer requires the installation of a dedicated Framework.


Gilles et Roland

This should please lordsmurf

[mcconnor]

Thank you for this software - looking forward to the next restoration task. On request of my family members - can it be made possible to minimize the program window while processing the tasks? Those can take quite some time and at the moment the main window can't be minimized or even moved after the process has started. And it stays on top of other programs making it hard to use the computer for hours....

[Steve Elliott]

It doesn't work. Can't find this codec, can't process this clip......load of shit.

[mcconnor]

It doesn't work. Can't find this codec, can't process this clip......load of shit.

I doubt even the highly involved developers can make anything of this short - and rude - error explanation. Perhaps you should consider a more friendly tone first...

[leadry]

Hi to all Film9 users and developers. I’ve been using this software for almost a year now as a part for restoring my own Super 8 films from 1966 to mid 80’s. I think it’s a fantastic way to use the quite advanced programs like VirtualDub and AviSynth. I use a RetroScan Universal 2k from Moviestuff for frame by frame capturing in 2048 x 1536 in uncompressed AVI and I use Sony Vegas as my NLE. I want to use the high resolution and uncompressed as long as possible in my work flow BUT one problem; I don’t seem to be able to run the Film9 software with 2048 x 1536 input and keep those dimensions for the output from Film9. The Film9 (VirtualDub) will crash quite immediately and say ”An exception occured in module KERNELBASE” among other things. Is that something to do with my PC or is it a limitation of the Film9 / VirtualDub. I use the latest version of the Film9 s/w.
Instead of keeping the 2k resolution I have to go down to 1440 x 1080 for the output file to make it work.
Any ideas from the experts of Film9 or VirtualDub?

Kurt

[GillesH]

Hello Kurt,
Sorry, but FILM9 will not work with this important resolution.
There is a limitation of calculation because of the number of pixels to be treated, even with a 64bits high performance PC.
We use 32bits versions (Avisynth + Plugins) and there are, necessarily, memory limitations.
VirtualDub2 has been optimized, but the overlay of Avisynth filters reaches its limits beyond Full HD.
And this can cause error messages like "KernelBase".
Gilles

[leadry]

Thank you so much for the explanation Gilles! It is still a great software the Film9 that you have created!

Best regards
Kurt

[Steve Elliott]

It WONT WORK!!!! I import the clip, adjust the setting and as soon as I click start processing I get two error messages. The first is about some nonsense about Net.something or other and the next is saying it can't find the clip I've attached. It's a waste of time. I don't suppose anybody will give a detailed blow by blow account of how they get it to work?

[GillesH]

Hi Steve,
It's amazing ! There are many people who say that it works very well.
Before using, please read the documentation.
And then, to get an idea of your problem, it should give us more precise information on your clip (MediaInfo), on the options of the Film9 Project, etc.
Possibly, a portion of clip download.
We can not guess what's happening to you ...

[johnmeyer]

.NET usually refers to the Microsoft .NET Framework runtime which is something that must be downloaded and installed. It is used by many programs. However, the installer for any program that requires its use should automatically do that download and installation for you. Here is a link to the Microsoft site where you can download it yourself:

Microsoft .NET Framework Download

[Gelinox]

.NET usually refers to the Microsoft .NET Framework runtime which is something that must be downloaded and installed. It is used by many programs. However, the installer for any program that requires its use should automatically do that download and installation for you. Here is a link to the Microsoft site where you can download it yourself:

Microsoft .NET Framework Download

Not always (Windows10 already has its built-in framework). Additionally some clip may cause this error.
As requested by Gilles, we need more details before proposing solutions.

[Kilohertz]

Hi all,

Getting to the party a little late but am starting to learn Film9 and VirtualDub. I have had a few crashes, nothing I'm worried about yet but may post once I know it's not something I am doing, but my question for today is...

Where does Film9 point to VirtualDub? i.e. Can I control which version of VirtualDub that I have installed on my computer, Film9 opens? Currently it is opening VirtualDub 2 build 43073. I have a few different versions loaded and would like to try them. I have looked in the config files in the Film9 directory but didn't see anything there. Guessing it's embedded somewhere in the .exe files??

Thanks!

okay I decided to investigate the crash, and it's a VirtualDub compression error, says may be corrupt data error -100, and it happens in any container using H.264, Lagarith is fine, lossless is fine. It crashes when it moves to the second pass. Interesting, frame count is 357 on first pass, but on the second pass it reports 594, FYI. Also I watched the log in VirtualDub as it crashed and it reported INPUT DECOMPRESSION YUV422, and OUTPUT COMPRESSION YUV422, another FYI

PS I really like the software, from what I have seen, it does an amazing job on old 8mm film.

[GillesH]

Hi Kilohertz,

FILM9 uses a version of Virtualdub2, but you can not change it.
In principle, this is one of latest version. Shekh (the designer of Vdub2) released a V43382 just recently.
This will be for the next update of FILM9.

Regarding the second point, there is probably a problem with Encoding Codec recognition.
For H264 encoding, we use the resources of Vdub2.
It is therefore necessary to use the version of Vdub2 which is included in FILM9 and not to try to modify anything.
If necessary, uninstall, clean and re-install FILM9.

However, you can use VirtualDub in parallel, but there is no link to create with the version of FILM9.

[Kilohertz]

Hello Gilles,

Thank you for the quick reply. Well I stayed up until midnight working on this, and I think I have figured it out, but won't know until I wake up a bit more and try it again. I was able to replicate the error, using just VirtualDub by itself, correct version 43073, and it has to do with the .264 codec. It has nothing to do with Film9. I watched the log and the error and it has something to do with the pixels not being square, or 1:1 or 2:2... I am going to investigate further today, trying different image capture types etc. I updated the codecs with newer versions, no difference. I was able to get the codec to work if I right clicked on the preview window within VirtualDub and changed the pixel size from "unknown" to 1:1.

My source files are a series of TIFF images captured from a machine vision camera (Allied Oscar 510C).

Thank you again for your work on this project. Speaking of next rev. I have a request. Is it possible to have more options for saving codecs and formats in the project settings?

Cheers

PS also reinstalled Film9 a few times but as mentioned, I don't think it is a Film9 problem

[GillesH]

Knowing that FILM9 is normally between a film scan and a NLE, we did not want to put too many Codecs.
It is especially the "non-destructive" Codecs that are to be preferred, even if the H264 can provide an alternative.
But you can specify the codecs you want. After, it will analyze their interest.
Best Regards

[Kilohertz]

Hello again,

Well I have been working on my telecine project for 3 weeks now and am about to start capturing and processing for real, and I have a few questions about how to best use Film9.

My work process so far is as follows, and I am open to suggestions and offers of help by any members more experience than I am.

I am capturing 8mm silent film with an Allied Vision machine vision camera, Oscar 510C, frame by frame, and capturing images that are debayered 8 bit files with resolution of 2588 x 1958 (native max resolution of camera). I had started saving them as BMP files but each file is 15MB, are uncompressed but take over 200GB for a 15 minute film. I also have saved some as TIFF files which are about 5MB file size, much better as far as space is concerned. As Film9 needs an actual movie file to process, I use VirtualDub2 64 bit 43382, and open the first file in the series of 17,000 or so files, then use the "save as" to create an AVI file which will be accepted by Film9. My first question is...what is the best codec to use at this stage, and what container is best to use with Film9, AVI, MP4 etc. As this is really my first serious foray into understanding codecs and video files, I could use some guidance on this point. I also noticed in post #252 that Gilles mentioned that Film9 can't work with large image sizes like this, so I am wondering if I should be saving in the max size that FIlm9 can work with? (please clarify what that size is)

So far I have been using Lagarith, which creates a rather large file, but it loads fine into Film9. I have managed to get Film9 to process these files, but with the 90GB file size, it takes over 9 hours to process, and then still generates a 50GB file. As Film9 is my final process (no further NLE) and I can't get it to save with H.264 (see above post, regarding it crashing on the second pass) I have been saving it again as an AVI Lagarith file, then once again opening that 50GB AVI file in VirutalDub 64 bit 43382 and saving it as another file, and again I could use some help here as I have no idea what format to save it as, bit depth, codec, container, pixel format etc. There is just a lot of items I just don't know how to set.

There are probably a lot of other things I need to ask, but I'll leave it for now as a first step.

Thank you for your help.

Cheers

[Gelinox]

Hello,

As you have an industrial camera, you can crop the image and choose the desired number of pixels.
So we will talk about a capture without unnecessary black borders. For a film of 8mm, 2588 x 1958 is a resolution much too big.
By keeping 1200 pixels maximum for the height, it's already very good.
To produce a movie in FullHd, it's more than enough. If the frame is clean and borderless, 1080 pixels are enough.

Then when you talk about frame-by-frame capture, I think you're using a trigger that automatically triggers the capture to ultimately get a video image for an 8mm movie frame.
My question is "why do not you capture a video directly?" Normally the software of an industrial camera should offer this choice.
And by reducing the definition, you will allow your camera to exceed the 3i / s that seem to be its maximum capture speed at full definition.

The following workflow must be obtained: Capture - Film9 - Editing program.

The crashes in .mp4 (H264) are due to an excessive definition of the image.
It seems that the limit supported by VirtualDub, (all the activated filters) corresponds to a bit more than a FullHD image dimension.Film9 should be used between capture and final editing. For this reason, it is best to choose an output format with Film9 other than .mp4(eg FFV1, lagarith, uncompressed format, AppleProres).

This is already a first step with many things to change …

[Kilohertz]

Hello,

As you have an industrial camera, you can crop the image and choose the desired number of pixels.
So we will talk about a capture without unnecessary black borders. For a film of 8mm, 2588 x 1958 is a resolution much too big.
By keeping 1200 pixels maximum for the height, it's already very good.
To produce a movie in FullHd, it's more than enough. If the frame is clean and borderless, 1080 pixels are enough.

From everything I have read, the goal is to achieve as much resolution as possible at capture so that no detail is lost in subsequent processing steps. I tried capturing at 1280x960, rearranged the camera and bellows and such to fill the frame at that resolution, and found I could see the pixelation, even on my computer monitor, so I went back to full native capture.

Then when you talk about frame-by-frame capture, I think you're using a trigger that automatically triggers the capture to ultimately get a video image for an 8mm movie frame. My question is "why do not you capture a video directly?" Normally the software of an industrial camera should offer this choice.
And by reducing the definition, you will allow your camera to exceed the 3i / s that seem to be its maximum capture speed at full definition.

I actually am capturing frame by frame, with a Hall effect sensor triggering the camera to capture a frame and record it to the hard drive as sequential TIFFs. I then use VirtualDub to make an AVI from all 17,000 TIFFs. The capture software does not have the ability to make an AVI.

The following workflow must be obtained: Capture - Film9 - Editing program.

As I am making no further edits after Film9 finishes, I am trying to make Film9 my final process, but I need to now figure out how to reduce the resolution so that it can save as H.264, or continue to use VirtualDub 64 bit 43382 to make the final compressed video, and having Film9 save as Prores or Lagarith as an interim step.

The crashes in .mp4 (H264) are due to an excessive definition of the image.
It seems that the limit supported by VirtualDub, (all the activated filters) corresponds to a bit more than a FullHD image dimension.Film9 should be used between capture and final editing. For this reason, it is best to choose an output format with Film9 other than .mp4(eg FFV1, lagarith, uncompressed format, AppleProres).

This is already a first step with many things to change …

Thanks for that explanation, it helps me understand the limitations.

I'll continue working on the process this weekend and post a video of my setup when I get a minute.

Cheers

[Gelinox]

From everything I have read, the goal is to achieve as much resolution as possible at capture so that no detail is lost in subsequent processing steps. I tried capturing at 1280x960, rearranged the camera and bellows and such to fill the frame at that resolution, and found I could see the pixelation, even on my computer monitor, so I went back to full native capture.

Not good ... The film has a definition. An 8mm image has a dimension of about 4.9 x 3.6 mm.
There is no point in wanting to use millions of pixels to scan an image that does not have that much. Unless you want to photograph the grain.

I actually am capturing frame by frame, with a Hall effect sensor triggering the camera to capture a frame and record it to the hard drive as sequential TIFFs. I then use VirtualDub to make an AVI from all 17,000 TIFFs. The capture software does not have the ability to make an AVI.

Change of software. See page 70 of your camera's instruction manual.
Capturing images, converting them, producing a video file, working it with Film9 .... that's not the method I would use.

But everyone does what they want, but I imagine that if you come to ask for advice, it is somewhere to take it into account and to improve the way you work.

[videofred]

I agree with Gelinox. High resolution is not always the best option for a good film transfer. It is better to have a good quality transfer at lower resolution than a bad quality transfer at high resolution. The latter is a waste of pixels and hard disk space.

I always capture at 1392x1032 with my Point Grey Flea and I am not missing anything. Film grain is very well captured.

It would be nice if Kiloherertz could upload some frames and/or clips to have an idea about the quality.

Fred.

[Kilohertz]

OKay boys, thank you for the input.

I will have another try at lower resolution and take it to the final step before I make a judgment. I have been battling with the Allied Vision software, trying to get it to work correctly, which I think I finally have done, but it has some limitations.

Which manual pg 70 are you referring to? The Oscar camera manual pg 70 speaks of debayering. Pg 70 of the SmartView manual speaks of streaming and recording. I have read all of these manuals cover to cover, including VimbaView which is another version of capture software from Allied. Last week, it was recommended to try AcquireControl, also from Allied. I have been working with one of their support engineers to get these software packages working correctly, it has been a struggle. My forehead is becoming flat.



If I move away from Allied software then I open up another chapter of learning, OpenCV or similar, I thought I was pretty good with computers but this project has taught me that I know very little about programming. As you mention, it would be best to capture the frames in an AVI file right away, rather than as TIFFs then converting to an AVI.

I will work on uploading a few clips for your evaluation today.

Thank you again.

[Gelinox]

Which manual pg 70 are you referring to? The Oscar camera manual pg 70 speaks of debayering. Pg 70 of the SmartView manual speaks of streaming and recording.

Pg 70 of the SmartView…


[Attachment 48264 - Click to enlarge]

Streampix is a software that should offer this possibility or at least a lot of possibilities (user manual 260 pages). It seems to be free from what I have seen.

[Kilohertz]

Thanks Gelinox,

Yes I saw that in the manual, which is why I can only capture sequential TIFFs with AVT software. I looked at Streampix and it's no longer free, VER7 only option is to request a quote. If someone has the older free version still that can post somewhere, I'd gladly download it, but at this point I already have over $1000 into the project...can't afford any more. Appreciate the link though.

So I have just spent the last 7 hours working on setting up my capture to be 1400 X 1050 and I have to say, it is much faster in every respect, but even after going thru to the final phase of Film9, I can tell it's HD, and it looks good on the TV, but I can see the camera pixels and that bothers me, whereas with full 2580 x 1958, I couln't see the pixels.

I'll try to get some video samples posted but my patience with computers is very low right now.

Cheers for now.

PS Actually I can post sample TIFF frames easily without going thru all the upload/download hassle of video, I'll do that right now.

This is full native resolution from a couple of days ago, raw capture from the camera, completely unprocessed. 2576 x 1958 saved as TIFF about 5M file size

https://www.dropbox.com/s/d02rg1ofllcgyhc/Film%208%20197200021.TIF?dl=0

This is a capture from today 1400 x 1050, same camera settings except for format. 2M file size. Still Format 7 Mode 4, just lower resolution. You can really see the pixelation in the basement window reflection.

https://www.dropbox.com/s/014izpx0zlw5h2d/TEST%201400%20X%201050%20RAW%208%20DEBAYERED00517.TIF?dl=0

[Kilohertz]

Gentlemen,

I must offer you my most sincere apology as I have discovered a whole new world of film capture and I am now only 20 minutes later getting GLORIOUS results.

After rethinking your recommendations, and looking at the TIFF captures, I knew there must be something wrong so I decided to "go wild" and reduce the resolution even further. Instead of using my old standby of Format 7 Mode 4, I just ran the projector and started playing with different "modes" in the camera. As I worked my way thru, I stumbled on F7 M2, which I had notes not to use as it may have side effects (something I read in the camera manual). Anyway, I landed on that mode and the image, although much smaller than what I had previously, looked darn good, and I couldn't see the pixels. So I reconfigured the camera and bellows again to get full image capture, reset the AOI and started just watching the frames as they rolled across the monitor. I couldn't believe it! ZERO pixels visible, but a WHOLE lot of film grain I hadn't seen before. The images looked good, it was a little harder to focus but got it dialed in and recorded over a minute of raw footage. The files are being saved as debayered 8 bit TIFF files, and are about 1.8M in size, resolution of 1280 x 968. I opened one and could not see the pixels at all, nothing like the full native capture files. So I ran thru the Film9 process, which only took 10 minutes for this 1.5 minute clip, and put it up on the big TV and WOW! It's amazing, smooth as glass, it blows away the captures I made in the last few days with full res images.

So there must be something to do with Format 7 Mode 4 that I don't understand, and will quiz AVT tech support about it, but for now it's great! I know I still have to make an AVI file out of all the TIFFs, but that is so quick now, and I use VirtualDub with no compression RGB and it loads perfectly into Film9. I am currently still saving the final file out of Film9 as an Apple ProRes, I need to try H.264 again, but the 1.5 minute movie is only 450MB and my PS3 plays it perfectly using Universal Media Server.

Anyway, a well deserved beer is awaiting me right now, I shall continue a bit later.

Thank you for pushing me to investigate this further, very much appreciated.

Cheers