Digitising my home video VHS tapes, confirm my method?
 

So I started trying to digitise my home video from the 90s and 00s back in 2018 and got overwhelmed by so may conflicting opinions.

Time has now passed so I thought I would have another go. I have read so so many forum guides and versions to do this and it's amazing how views change over the years and how many opinions there are. So I have pulled at the best ideas (I recon) and going to follow this process

So here goes and I'm putting this up as a guide to what and how Im doing it. So if Im making a real error here someone may have an opinion. I am not an expert or even a novice. Just a guy trying to capture his old films :)

I dont expect to get this done at the best 100% quality, just enough that we (i.e: family) can watch them again as and when we like

I'll be capturing VHS tape and also old camcorder tapes - being played on a VHS player (not the camcorder - it's dead) (UK - PAL) and I have an Adapter to put the camcorder tapes into as well as standard VHS tapes

I am using:
- WIN 10 64 bit
- Panasonic video player
- Dazzle DVC100 capture device
- VirtualDub (Capture)
- VirtualDub2 (Edit and Export)

- Using Lagarith (version 1.3.27) to capture, I am not using HuffYUY as I have read on this forum it has issues on Windows 10, which was proven right when I did a capture and the audio was out of sync with the video. I swapped to Lagarith and it was fine
- not using VirtualDub2 to capture - Lord Smurf said in a post to capture on standard VirtualDub not VirtualDub2. Therefore, I will be using VirtualDub2 to edit and compress

There will be 3 parts to my capture process:
1. The original capture that will be archived as a master
2. A version from the master - tidied up and compressed to watch whenever
3. A version from the master - tidied up, deinterlaced and compressed and uploaded to YouTube


Step 1 Capturing

1. Open VirtualDub and enter capture mode

2. I'll be using all the settings in this guide (previously recommended) unless otherwise changed below -
https://www.digitalfaq.com/forum/vid...-settings.html

From the Virtualdub windows menu across the top.......
3. Device / Dazzle DVC100

4. Video / Preview

5. Video / Capture Pin:
Attachment 16122

6. Video / Capture Filter:
Attachment 16123

7. Video / Compression
Attachment 16124
Attachment 16125

8. Capture / Timing
Attachment 16126

9. Then start capturing :)



Step 2 Tidying and archiving to watch (keeping it as unmolested as possible)

1. In virtualDub2 open the master capture file

2. Video / Filters - Masking the noise from the edges - using the guide as suggested by Lord Smurf https://www.digitalfaq.com/forum/vid...erly-crop.html

3. and selecting resize Setting
Attachment 16127

4. Then Cropping the noise from the edges
Attachment 16128

5. Video/Compression
Attachment 16129

6. Configure
Attachment 16130

7. Audio/Compression
Attachment 16131

8. File/Save Video (MP4)
Attachment 16132



Step 3 - These additional steps are for the Tube version - steps 1 to 4 as above - plus:
1. Add DeInterlace
Attachment 16133

2. Resize to 1440 x 1080
Attachment 16134

3.Then steps 5 to 7 as above

4. File/Save Video (MP4)


Personally on viewing them back I think the YouTube (1080) version may look better than the tidying (interlaced) version. Not sure why. So before I commit to recording and editing hours of footage I thought I would get some thoughts and views.

I tried to keep this guide as simple as poss and keep the whole process simple.

Raw capture https://drive.google.com/file/d/1nX7...usp=share_link
Tidied version https://drive.google.com/file/d/1YRK...usp=share_link
Tidied Youtube version https://www.youtube.com/watch?v=HPLU8JuojWA

Which Panasonic?
Some quality, like AG-1980?
Or a low-end PV model sold for watching rented tapes from Blockbuster?

Those timing settings can be fine. Or not. "Auto disable resync" is needed for cards. A tiny handful of cards need to disable preview timestamps. Some of that depends on the OS, and exact system, so it's not rigid rules. Just something to watch for, which is seen/heard with drops or audio attenuation (chipmunk, baritone).

Converting to Youtube with VirtualDub works, but nowhere as good as using Hybrid. For starters, Hybrid has QTGMC deinterlace, not ancient Yadif.

You must crop after deinterlace, not before! :warning:

That's not a good capture card, often very destructive. I have a DVC100 unit that I'd consider minimally passasble, but it seems that most are simply destructive. The values are all over the place, and it too often blows out the signal (hot, overexposed). The AGC inside is crap.

You lack any TBC, which is why the video is so degraded. Those wiggly train tracks should be straight, not a mushy artifact-filled scene. The rough Yadif makes it worse. At very least, add ES10/15. Right now, this video looks like it's being viewed with binoculars through a dirty window screen. With proper capture, it can look more like a retail DVD.

Reply to post
 

Thanks so much for the reply and help :)

Panasonic NV-VP31

Ok thanks will tick the box and try it out

I read Hybrid was not for newbies so I never tried it. I'll look into it - unless you can point me to a post that has a guide

Ok will do - but above you say deinterlace with Hybrid. So, do you also crop back in Virtualdub after deinterlacing in Hybrid.
P.S: in a previous post you said "never crop - always mask". Which is what Im doing in both outputs - is that still correct :)

:warning:

Fair enough - but it's all I got :)

I dont have a TBC - trying to do this on a budget as I dont have that many tapes to do - just not cost affective.

Also - what is "add ES10/15"

And I connected up the video player to a TV and in all fairness the train tracks still look bad. This was my Father in Laws tape so I cant vouch for the camera he used to record it. I'll try a different recording and repost tomorrow

Hybrid isn't "not for newbies". Anybody can learn anything. But it requires some reading, often some hands-on testing. Not blind pressing buttons like an ape. Inversely, don't think you're an instant expert because you got a result that didn't suck. Use it, go slow, proceed with caution, learn. If you can find your way through the maze that is encoding x264/5 in VirtualDub2, then Hybrid isn't a huge leap. It has defaults, but the key to quality is learn what nuanced settings do, not really much different from VirtualDub2 use of x264vfw, though it didn't show all the settings like Hybrid does.

DMR-ES10, DRM15. Read the forum for info on ES10/15 units, as a very basic TBC(ish) for passthrough. It's not ideal, not a TBC replacement of any sort, but vastly better than nothing at all.

Being on a budget is fine. But realize there's a difference between intelligent budgeting, and being too cheap. Not buying something isn't budgeting, it's just being cheap. Some form of TBC is required, even the inexpensive ES10/15 as the TBC(ish). Dropped frames on your video is obvious, and quality stinks. Again, ES10/15 is a bare minimum, not flawless. You'll see an obvious boost by using it, just realize it gets better or more stable. Buy the ES10/15.

You can get other cards, for not much cost. Certain PAL users here like the Hauppauge Live2. I don't like this model, but it would be a big improvement here, paired with ES10/15. Right now, you're making mushy crummy videos, but this small change will vastly improve it.

The VCR, and lack of TBC, is why the tape is bad. It likely has nothing to do with the signal on the original tape, unless it's nth gen (and if that's the tape he shot, it's not nth gen, aka a copy of a copy of a copy etc).

Ok downloaded it and giving it a whirl - any good site for recommended guidance?

Ok I've bought a DMR-ES10 - it's on the way - I have never used a TBC before - I presume you go VHS out to ES10 in then ES10 out to capture card to PC


If not a Hauppauge Live2, what would be better?

So I received my Panasonic DMR-ES10

I recorded the same footage as above, but this time I:
1. Recorded the RAW footage
2. Used Hybrid to De-interlace using Avisynth (QTGNMC) and bob and made it 50 FPS
3. Then used Virtualdub2 to crop and letterbox to remove noise and resize to 1440X1152
4. Then went back to Hybrid to recode back to x264
5. Uploaded to Youtube

This is the result

New TBC and Hybrid version https://www.youtube.com/watch?v=dtctNT_7qCo

compared to a non TBC (ish):

Tidied Youtube version https://www.youtube.com/watch?v=HPLU8JuojWA

It should be 1440x1080 not 1440X1152, Since youtube takes only square pixel, it has probably already been squeezed back to 1080 which screws up the frame geometry, I don't have access to YT from here but I will check back later.

Youtube will display, whatever you give it if they are square pixels; it doesn't stick to "norms". In this case, YT is displaying 1440x1152 as 1.25, which it is.

Why not just use the x264 8 bit encoder in VDub2? More a question for the floor, I guess. One less encode (admittedly, lossless, I assume).

reply
 

I only did it as 1440x1152 because I captured in PAL 720x576. I did a lot of research yesterday and apart from the guide in my original post at the top for going 1080 for YT. I cant see a good option to resize to 1080 with chopping away pixels on the side. Unless anyone can suggest a post to go from 720x576 to 1440x1080

So using Hybrid I want to do all the following in Hybrid:
- Deinterlace
- crop the noise and add a frame around the edges to get back to 720x576
- then resize to 1440x1080
- then encode using x264

I could not find a way using Hybrid to add a frame like you can in Virtualdub2. Hence I then went between the Hybrid and VirtualDub2. Unless you can add a frame like VirtualDub2


I have found that encoding the attached video. In Hybrid the size was only 15MB, but in VurtualDub2 is was encoding at 38MB - is probably a settings thing - but I am not experienced and a newbie to this - so learning as I go

720x576 -> 704x576 -> 1440x1080 This is the correct aspect ratio,
Or 720x576 -> Crop to the active video area with clean edges but try to stay within 704:576 ratio, for example 698x572 then resize to 1440x1080, 6 pixels off give or take is not going to be noticeable.

No, Youtube always defaults to 1080, I just downloaded the video and it is 1080.

I have now done the following:

Deinterlaced in Hybrid using Avisynth and bob to 50 fps and 8000Kbits per second - encoded using X264 - saved as MP4 - size was 71.5MB

Removed Noise on the sides of the video in Virtualdub2 and framed then upscaled to 1440x1080 - saved as MP4 - size was 16.1GB

Back in Hybrid and encoded again using X264 - size went down to 73MB

result was.... https://www.youtube.com/watch?v=iPYs2GAYuck

I will record a different video and follow this same process and upload here, as I dont like the existing clip as I think the original quality is poor

Yes, it is confusing. The issue is that although a capture is only 720x576 (5:4, or 1.25 to 1) it is designed to be displayed at 4:3, or 1.3333. to 1. While it seems all SD is 720x576, it only looks correct because the pixels are stretched a bit. The effective ratio is 768x576, which is also 4:3, or 1.333:1.

Thankfully, the world came to it's senses and virtually every other image size above SD is worked out using square pixels, so a 4:3 video (which is what the effective/apparent size of SD is) equates to 1440x1080. What I think you did was, completely understandably, apply the 720x576 ratio to the 1440 frame size to arrive at 1152.

Now, how to.

First, if you are going to resize to 1440x1080, you must de-interlace first. VDub has a deinterlace filter which works pretty well. You can (just) tell the difference between it and QTGMC but is it worth the hassle going to Hybrid for your deinterlacing then back to VDub for cropping and resizing? If it's a super-important video for wide distribution then maybe yes, but for my family stuff, nope.

In VDub2 (I use 2 because it has a proper cropping filter as opposed to the "addon" that 1.9.11 has, as well as the native x264 export options), open up the crop filter.

Crop off the vertical edges as desired. However much you cropped off the top and bottom, multiply it by 1.333, then add 16, to give you the total to be cropped off the sides (the 16 chops off the "dead area" that is on the sides of the captured frame but which doesn't contain any video). If you need to take more off the sides, just adjust the vertical crop to match the 4:3 ratio.

Example: crop 15 pixels off the bottom to get rid of the VHS switching noise. Multiply that by 1.3333 (standard SD ratio) to give you 20. Add 16 to that to give you 36, and that's the total amount you have to crop off bottom and top.

Finally, set up the resize: go to the Resize filter; in the Aspect Ratio area, tick the "Disabled" radio button. In the New Size area, tick "Absolute" and enter 1440x1080 (or 768x576 if wanted). OK out.

On the audio menu, choose Full Processing.

For the "Save video...", in the x264 Configure, set the SAR to 1 and 1 and set the other parameters as you need. If not setting an average bitrate, set the CRF to something like 16 and play with it. Lower numbers give better quality but bigger file sizes.

Then back to the Save As screen, set the file extension to MP4 Fast Start and audio compression to FFMPEG AAC.

Done!

You obviously didn't notice the Youtube horizontal size is 1350, which is 1.25:1, same as 1440x1152, and which is the ratio that YT displays the video at, as per the screen grab attached.

@Jayce72, try using the deinterlace filter in VDub: choose Yadif and double frame rate TFF. :wink2:

reply
 

Thanks and why does virtualdub2 encoding in x264 create a file size at 16GB compared to Hybrid at 78MB - thats a huge difference

reply
 

I am deinterlacing in Hybrid off the advise from Lord Smurf that Yadid was now quite old and QTGMC was now far better and Hybrid was now better at encoding

I like the fact YouTube added a "stats for nerds" option (right-click on video). This also shows the video is 1440x1152:

It is indeed a huge difference. It'll be the bitrates, unless you are mistakenly exporting uncompressed video in the second step out of VDub (as opposed to H264).

If you insist (at Lord's Smurf's insistence :laugh:) on going via Hybrid and back, you should try to not go via the two in a lossy codec. You're currently re-encoding MP4 twice that I can see. This is not good, especially for SD.

I suggest this workflow:

Open your file in Hybrid and deinterlace.

Export from Hybrid using a lossless codec such as HUFFYUV or FFV1 (unfortunately, Hybrid doesn't support LAGS in my version).

Open that in VDub, do your stuff, then re-export as another lossless (HUFFYUV, LAGS or FFV1).

Open in Hybrid and set up your final MP4 export.

That workflow will keep generation loss to a minimum because you are working with lossless files until the end.

I had a quick play around with Hybrid (I gave up on it previously, doing what you are doing, because of the slow encode speed for QTGMC. I confess that I now use QTGMC for my high-value files; although it did take 5 years off my life to get it working!).

Anyway, it looks to me like you can crop and resize in Hybrid: you can click on "crop view" at the bottom right of the Crop screen. I haven't fully worked it out but it seemed to be working; it pops up a preview.

You can also set up a resize there, using the same principle as I described above. For 1440x1080, the final SAR/DAR/PAR whatever it is is Custom 1:1 square. :smack:

You may be able to do the whole thing in Hybrid.

Neither are as clear or simple as in VDub but it will certainly save a re-encode or two. If you're doing a proper-sized video and not a couple of minute old TV ad, these re-encodes take time.

I forgot about that. Thanks for the reminder.

And while you're at it, do a comparison between Hybrid/QTGMC and VDub/Yadif/Double FR. Make sure the bitrates and codecs are the same.

Another thing: Mediainfo is a great little program for analysing files. It tells you everything there is to know and is especially helpful for issues such as the 16gb export you did. One caveat: captured AVI video (eg 720x576) does not have an aspect ratio flag in the file header, so Mediainfo (and other programs) simply divide the width by the height to come up with 5:4. The video is, of course, 4:3.

A new train set is already ordered and slowly being built. :cool:

Yes, youtube algorithm doesn't care about width, but it always defaults to 1080p as the standard vertical resolution and re-calculate the horizontal resolution assuming square pixel, If you don't lock the aspect ratio youtube will stretch the frame because it treat them as square, You can clearly see that his non standard resolution stretched the video.

So I have recorded some new footage using the Panasonic DMR ES-10 as a TBC and audio direct to capture card

I have used VirtualDub2 to deinterlace using Yadiff and also to crop and resize.

I have uploaded 4 versions (all are deinterlaced)

1. is just 720x576 https://www.youtube.com/watch?v=n5JHFW34cyw
2. is the 720x576 stretched to 1440x1080 https://youtu.be/O90WoLJCeQU
3. Is the 720x576 resized to 1440x1152 https://youtu.be/EkyKxfbILxU
4. Is the 720x576 cropped to 4:3 and resized to 1440x1080 https://youtu.be/Nm5ZIm16gSk

Good job ! but...

Video 1 (720x576) is not 4:3. It's 1.25:1 (or 5:4). Have a look at/measure with a ruler at the physical YT dimensions on your screen (I use a screen capture program which shows the pixel height and width).

After you cropped it, if you're going for square pixels (the simplest in my mind) you resize it to 768x576. That's 4:3. Otherwise, you have to export it using one of those SARs (16:15?? I think for PAL SD).

Video 2 corrects the error in video 1 because 1440x1080 is 4:3 (1.33333 to 1), just like 768x576.

Video 3 shouldn't be 1152 because it stretches the video (1440x1152 is 1.25:1). That will be evident if you measure the YT screen size.

Video 4 is good! Similar result to video 2.

Whatever you do with resizing and export SARs, the end result must be a 4:3, or 1.33333 to 1 display on the Youtube screen.

I'd apply a bit of light denoising. I find the Virtual Dub MSU Denoiser v 251 works pretty well for the (free) price. I get best results using the preset "Medium Noise MSU" and both sliders on 42. That will remove most of that colour blotchiness (and the small cost of sharpness).

Video #2 have like 1/2 a second, But all have wrong aspect ratio, You failed to understand that you should crop with a ratio of 704:576 before you resize to 1440x1080, I don't see any point of the resolution 1440x1152. Good luck.

Could you please remind us where the "magic" number 704 comes from and why that value ?
Thx

So, now:

Using virtualdub2
- Deinterlaced using QTGMC
- Resized to 1440x1080 (from PAL 720x576)
- Letterboxed framing to remove noise
- 4:3 aspect ratio
- Compressed x264
- Saved as MP4

https://www.youtube.com/watch?v=YhYPkjpbytA

Looks good to me!

Putting on my heretic's hat, I don't subscribe to this masking stuff, especially if they are not symmetrical. I think it looks amateurish. Few YT videos are masked; they are cropped. The video is going to be re-encoded anyway so you're not "losing" quality by dropping as opposed to masking.

I don't think masking is amateur at all. What I find truly amateur is when SD or vertical phone footage is repeated and blurred behind the video, to fill in the black bars. People need to get over their obsession (OCD) with "black bars" and "black outlines". The same folks are often the ones who watch 4x3 as 16x9.

There are times to crop, and times to mask. It's always about the destruction of the content. Some footage cannot withstand cropping (and thus enlargement) without detriment. I'm not 100% either way, but always defer to the safe method of masking. You can always mask fine, you cannot always crop fine.