Suggestions requested to optimize restoration of my LaserDiscs in StaxRip

[SteveLarson]

I have a LaserDisc collection of about 75 titles and have been capturing a number of them in OBS Studio via the Hauppauge 610 Video Capture Device connected to my Pioneer CLD-V2800 player and ASUS ProArt P16 laptop. It took me a year of experimenting with various settings in OBS after recording lots of discs, but I finally settled on a group of core settings that yield me a completed capture I am pleased with. All my discs are NTSC and live action, although I plan to add several animated and anime titles as well. I’ve been capturing them at a resolution of 640x480 and with a color space of Rec. 601 (limited range of colors). While the preset fps in my Hauppauge settings is 29.97, I set the Common FPS Values for the output in OBS’ settings to 59.94 fps after reading posts that frames actually get dropped or deleted if captured at lower frame rates. I would rather continue capturing at 59.94 or 60 for the output because I don’t want to lose any frames. OBS automatically converts all interlaced content to progressive. I’ve been recording both the analog and stereo mixes in PCM. The majority of them sound excellent.

Throughout the second half of this year, I’ve been researching and learning about many filters and scripts in AviSynth, which I want to use for much of my post-processing work. I have StaxRip (version 2.50.2-x64) and many plugins/filters from the archived package, which also include those for VapourSynth that I downloaded from GitHub. I’ve been assembling a tentative plan for my workflow and would especially welcome your suggestions to tweak or perhaps add different filters. Feedback on parameters and preset values is also much appreciated.

I have watched at least 60 of my LDs and want the filters to take care of the most common global issues associated with this analog format: chroma noise, jagged edges/shimmering, dot crawl, rainbow patterns, and random video artifacts that pop up. (I may engage in more extensive frame-by-frame restorations for certain titles.) While I would like to clean up those anomalies to a considerable degree, I want to preserve the original texture and detail of the image as much as possible. I have envisioned my restoration project in three parts:

(1) Since nearly every one of my titles was originally shot on film before transfer to LD, I will perform an inverse telecine for each. Since there may have been duplicate frames created during capture, I will make sure TDecimate is part of my script. All of this goes towards converting 59.94 to 23.976.

(2) Anti-aliasing, very moderate noise reduction, and restoration cleanup.

(3) I will be eventually transferring the finished file to a BD-R (720p), which I’ll watch on my QLED TV. I have a region-free Sony Blu-ray player that has an option to output precisely at 720p. At this stage, I will want to convert YUV to RGB and Rec. 601 to Rec. 709. I will be upscaling/resizing the captured SD files to 960x720.

Here are my scripts and projected order for each job’s workflow:

Part 1

Field:
TIVTC
TFM().TDecimate()
cthresh=255

Frame Rate:
ChangeFPS(24000, 1001)


Part 2

Line:
DAA
daa3mod()

Noise:
mClean(thSAD=400, chroma=true, sharp=10, rn=14, deband=4, depth=0, strength=20)

Restoration:
ChubbyRain2(th=10, radius=10, sft=10) # progressive sources only
DOTKILLS(ITERATIONS=1)

Note: I also will probably use DeScratch to minimize and/or remove scratches/tramlines from B&W movies and older films shot in color.


Part 3

Color: YUV / RGB
YUV to RGB
Dither_convert_yuv_to_rgb()
ColorMatrix(mode="rec.601->rec.709")

I will be using ffmpeg - ProRes as my encoder.
Encoder command line:
Custom: --colorprim bt709 --transfer bt709 --colormatrix bt709

MOV is my container.

Spline64Resize(960, 720, 0, 0.5)

[Alwyn]

If you're doing that much restoration, the strong recommendation from most of us would be to not capture with OBS. QTGMC will do a much better job than OBS at deinterlacing your captures. A lossless capture with the Live2 (using AmarecTV or VDub) should not result in dropped frames and not require capturing at 59.94.

OBS is OK if you just want a one-and-done capture in reasonable quality, but in your case it seems you're going to do much more.

Regarding playing them on your TV, a portable USB drive would be more convenient than having to burn BDs.

Out of interest, what video codec are you capturing into?

[The 14th Doctor]

Since you are also dealing with LaserDisc, if you want the best quality possible then you might want to look using a Domesday Duplicator+LasterDisc Decode method, unlike it's younger offspring VHS-Decode, this method is 100% the best and proper way to archive LaserDiscs at their highest possible quality.

Since it bypasses the crappy composite connection used in LD players altogether.

There is however a massive learning curve, so be warned..

[SteveLarson]

If you're doing that much restoration, the strong recommendation from most of us would be to not capture with OBS. QTGMC will do a much better job than OBS at deinterlacing your captures. A lossless capture with the Live2 (using AmarecTV or VDub) should not result in dropped frames and not require capturing at 59.94.

OBS is OK if you just want a one-and-done capture in reasonable quality, but in your case it seems you're going to do much more.

Regarding playing them on your TV, a portable USB drive would be more convenient than having to burn BDs.

Out of interest, what video codec are you capturing into?

Thanks for all that feedback! I capture into AMD HW H.264 (AVC). My container is MKV.

I've actually never used AmarecTV. Have you and how well does it work? Had good results?

I have used QTGMC in StaxRip and it's been great for my interlaced videos.

[Alwyn]

I capture into AMD HW H.264 (AVC). My container is MKV.

OK, AVC isn't recommended for extensive restoration work.

I've actually never used AmarecTV. Have you and how well does it work? Had good results?

Yes. Here are my instructions.

This topic will probably get some pushback re AmarecTV, but in my experience, it just works. The other option is VDub. The VDub fanboys can guide you on that.

Both proggies produce 29.97 interlaced, lossless AVI files which deinterlace with QTGMC really well, and of course lend themselves to other (your) restoration techniques, including IVTC. A recommended free AVI codec is UTVideo (4cc ULY2).

[lollo]

This topic will probably get some pushback re AmarecTV

It should not, because there are no concrete reasons for that. Concerning reported dropped and inserted frames some post-processing of the AmarecTV report should be done to fully understand its mechanism. I have some interesting example if you wish.

To OP:

• Hauppauge USB-Live 2 is not the best for capturing from a composite source, because its poor comb filter. Introduce a Y/C generated signal in the path or change card for better quality
• Use AmarecTV, not OBS
• Capture 720x480 YUY2 4:2:2 lossless
• Your sources are probably not interlaced but telecined, so no QTGMC should be needed
• No RGB conversion is needed
• Staxrip is just a GUI, go to direct AviSynth processing
• Domesday Duplicator is the best method for LDs archiving

[Sharc]

^^^^
This, +1. And if you stick to the 610 USB-Live2 capture device use a recommended DVD recorder (such as a Panasonic ES-10 or ES-15) in passthrough which will add its Y/C separating comb filter to the workflow and strongly mitigates the typical composite issues (dotcrawl and rainbows).
Plenty of posts about this in this forum.

[SteveLarson]

Thanks to all for your insights and useful suggestions. I want to reply to each of your points when I get a chance. I found a Japanese-made capture device which, according to at least one person who's used it, has an excellent comb filter. So I'll probably be picking that up. I downloaded and tested AmarecTV last night with some samples from my Timecop LD. I love that the capture/viewing window is a lot larger than the one in OBS. By coincidence, I searched and found the user guide on AmarecTV that Alwyn put together. Quite helpful. I am very meticulous about getting the configuration settings exactly right. I may have some follow ?s after I obtain the new capture card.

[Sharc]

I found a Japanese-made capture device which, according to at least one person who's used it, has an excellent comb filter.

Mind you telling us your finding? Hoping you are right ....

[lollo]

It's probably the IO Data GV-USB2

Its problem is that in its HQ version and with the new drivers tends to insert a lot of frames for VHS/S-VHS signals otherwise behaving properly with other cards (in my experiments).

I ignore how it behaves with the "clean" laserdisc input signal, because I captured mine sourced by a high-end LD player with Y/C output using the USB-Live 2.

[Alwyn]

@Lollo, have you tried with the "older" drivers?

[lollo]

Yes, I tried the following on 4 different tapes:

HP laptop, Windows 7, drivers 1.1.0.93, AmarecTV3.10, VirtualDub 1.9.11

Asus laptop, Windows 11, drivers 1.1.0.193, AmarecTV3.10, VirtualDub 1.9.11, VirtualDub2 (latest version, just for fun)

Asus laptop, Windows 11, drivers 1.1.0.199, AmarecTV3.10, VirtualDub 1.9.11, VirtualDub2 (latest version, just for fun)

all the same results, between 10 and 22 inserted frames across 1 hour capture.

All the inserted frames are caused by previous dropped frames, and then in the AmarecTV log file (the only program providing a detailed report on what happens during capture) are reported just as inserted, while the previous drops are not reported. But they are easily found with a basic post-processing of the log file, checking the incoming video frame (VT=...) and the captured frames (Cap=...) (NT=...).
Linked here an example of a recent GV-USB2 capture versus two USB-Live 2 captures of the same tape done in 2021 and 2023: https://docs.google.com/spreadsheets/d/1duZgh2ns829CxyKH7pCabsyIJMWSE8VU/edit?usp=shar...f=true&sd=true

Possible reasons:
- faulty card [unlikely]
- Asus PC Windows 11 not performing well (but another PC with Windows 7 shows same behaviour) [unlikely]
- tape has degraded across the years and the captures I made (across multiple captures the inserted frames appear most of the time at the same instant) [most likely]
- GV-USB not dealing well with frames not 100% clean (I noticed that GV-USB drops frames that have some disturbance, later inserting a "repetition" frame, while USB-Live 2 and other cards capture these without problem [possible]

To remove any doubt I should now run a new capture with a different card and the different combinations of hardware and software. Will do later

[Sharc]

..... and then in the AmarecTV log file (the only program providing a detailed report on what happens during capture) are reported just as inserted, while the previous drops are not reported.

Confirmed. The indication 'Drops 0' is unfortunately misleading without detailed post analysis of the log.

Linked here an example of a recent GV-USB2 capture versus two USB-Live 2 captures of the same tape done in 2021 and 2023...
Possible reasons ...

Were the tapes played with the same VCR with the same VCR settings (TBC, DNR ...) for the comparisons? Also consider that besides the tapes the VCR heads and mechanics are probably in different condition between the captures of 2021 and now.

[lollo]

Yes Sharc, AmarecTV reports dropped frames non compensated by later inserted frames in the field D(-). But it’s ok, because in the log you have all needed infos. Also Buf(=…) info is useful for debugging. Still the best capturing software!

The same VCR with same setting is used. What I found interesting is that GV-USB2 video chip refuses to capture the frames that are somehow partially corrupted on the tape since the original recording. It is somehow less resilient, at least the HQ version I use now (but I assume HQ suffix is just for different software available)

[Sharc]

The same VCR with same setting is used. What I found interesting is that GV-USB2 video chip refuses to capture the frames that are somehow partially corrupted on the tape since the original recording. It is somehow less resilient, at least the HQ version I use now (but I assume HQ suffix is just for different software available)

Interesting finding. I have not noticed it and can neither confirm nor deny it, probably because I often had the ES-15 (or equivalent) in the loop which may have masked such resilience deficiency of the capture device.

[Alwyn]

@Lollo, what's your hardware workflow? I've never seen 10-20 drops/inserts per hour with my GV on the 2010 and 2021 drivers.

[lollo]

I have not noticed it and can neither confirm nor deny it, probably because I often had the ES-15 (or equivalent) in the loop which may have masked such resilience deficiency of the capture device.

Yes, ES-15 or equivalent may discard "marginal" frames, so you dot not even see them. You can recognize a drop only if there is an evident jump in motion or if you have a trusted source to compare (a DVD release or a DVB-S/T recording of the same video).

@Lollo, what's your hardware workflow?

Very simple: JVC HR-S9500MS Y/C output to IO Data GV-USB with high-end Monster Cables for audio and video. Capturing as 720x576 4:2:2 interlaced lossless HuffYUV and uncompressed audio.

I've never seen 10-20 drops/inserts per hour with my GV on the 2010 and 2021 drivers.

Me neither until now . It may just be a specific problem on my side with tape degradation and/or new hardware (new Windos11 PC and new GV-USB2 HQ).

[Sharc]

I have not noticed it and can neither confirm nor deny it, probably because I often had the ES-15 (or equivalent) in the loop which may have masked such resilience deficiency of the capture device.

Yes, ES-15 or equivalent may discard "marginal" frames, so you dot not even see them.

In this case one would expect "orphaned" inserted frames by Amarec to keep A/V in sync, means without corresponding reported dropped frames.
(Depending on motion it can be difficult to reliably identify dropped frames visually).

[lollo]

In this case one would expect "orphaned" inserted frames by Amarec to keep A/V in sync, means without corresponding reported dropped frames.

Yes!

[johnmeyer]

Depending on motion it can be difficult to reliably identify dropped frames visually.

True. However, I did make an attempt which actually (sort of) works:

Automatically fix dups followed (eventually) by drops

It is actually fairly useful but does suffer from a problem I was never able to solve, namely that TDecimate uses a fixed window for deciding which frames to decimate, and if the duplicate frames fall into adjacent windows, rather within the fixed window, they will not be correctly decimated. This problem can be reduced by making the windows size really large, but this slows down the script while not completely fixing the problem. I did ask for help, but was never able to come up with a satisfactory solution.

However, in many cases, you can reduce the problem by 90% or more by running the script.

[Sharc]

Depending on motion it can be difficult to reliably identify dropped frames visually.

True. However, I did make an attempt which actually (sort of) works:

Automatically fix dups followed (eventually) by drops

It is actually fairly useful but does suffer from a problem I was never able to solve, namely that TDecimate uses a fixed window for deciding which frames to decimate, and if the duplicate frames fall into adjacent windows, rather within the fixed window, they will not be correctly decimated. This problem can be reduced by making the windows size really large, but this slows down the script while not completely fixing the problem. I did ask for help, but was never able to come up with a satisfactory solution.

However, in many cases, you can reduce the problem by 90% or more by running the script.

Thanks John. Yes I have been using your script (or hello_hello's modified FixJumps3b.avsi) in the past with some success. I added RIFE as the interoplating algo, but of course this didn't address the pending issue of the fixed window size. Anyway, it worked nicely on a friend's ~18fps film source which was perfectly scanned but awkwardly converted to 25fps video and edited, leaving me with blends, duplicates and drops

[Alwyn]

@Lollo, chuck a stabiliser in the flow to see what happens. The Live2 has been less tolerant for me compared to the GV, as I have pointed out before.