Editing VHS captures with Sony Vegas

[VHS Roger]

The source is actually Video8/Hi8 (not sure which). I used VirtualDub with HuffYUV codec for capturing. I am a newbie with video capturing/editing so I don't know the best method for editing the videos. I am using Sony Vegas for editing. I included a sample clip to show the original capture and after I've edited. There are several things I'm not sure about:

• I seem to get a sort of watery/ripple look sometimes after editing.
  
• I don't know if I should worry about going between YUV and RGB.
  
• I generally need tips on making the video look as good as it can (color curves/levels etc).

Thanks for the help!

[LMotlow]

Your clip.avi is RGB24. Did you capture to RGB? Don't. If you cut this in VirtualDub for a sample and saved with "full processing mode", your changed the original colorspace if it wasn't RGB. Don't.

Your Clip Edited.avi has been deinterlaced (why?). Improper deinterlacing (alternate fields discarded). Don't. You lost 50% of your video by doing so and created motion problems. What do you want for final delivery format and family playback? DVD? SD-BluRay? AVCHD? Those formats are interlaced in standard definition video.

Your original capture sample has excessive interlace combing, typical for video8/hi8. Not unusual. Also dot crawl. There are ways to fix it without destroying half of your video. But not in Vegas or Virtualdub.

Cute kids.

[VHS Roger]

AFAIK, I captured in YUV. I tried cutting again using "fast recompress" and "normal recompress", but both created a very small file with no video. I used "Direct Stream Copy" and that video is linked below.

I am editing this video to simply be viewed on a computer, no DVD or BluRay. I tried deinterlacing in Vegas because it seems to look better, but I'm not sure I'm using the best settings.

[VHS Roger]

Forgot to attach the video...

[LMotlow]

Thank you, the YUY2 capture has cleaner color. The usual tape noise aside, there are easily visible artifacts every time something moves, and dot crawl really gets in the way. Could you give us some details about how your 8's were captured? What capture device and software was used? How did you play the tapes? How was the player connected to the capture setup (composite cable, s-video, etc.)?

[VHS Roger]

I used a Sony NTSC EV-C100 Hi8 player to play the tapes. I captured with a Diamond VC500 One-Touch Video Capture device connected via S-Video. I used the "Capture AVI" in VirtualDub and HuffYUV codec.

I've captured 50+ hours of tapes, so unfortunately re-capturing isn't really an option even if I didn't use the best method.

[LMotlow]

Thanks again. The more troublesome aspects look like playback problems to me. If all of your captures have similar problems, then playback is where is starts. The VC500 isn't the cream of the crop (and who can afford the cream?), but it's a pretty decent card for tape capture and far better than the $20 wonders being hawked everywhere. So I doubt that the VC500 is at fault. It's possible, but I'd say it's the player that needs maintenance.

In any case, we've seen worse -- much worse. You're a step ahead with lossless media, even if it does need some repair work. You should be aware that most NLE editors like Vegas, Premiere Pro, etc., are just that -- editors. They can cut and join, apply special effects and compositing work, encode, and do some sophisticated color correction. But they're not good choices for repair and restoration. Your best friend here will be Avisynth, without getting 'way too complicated, and you can get further tweaks with VirtualDub.

One of the glitches you might not have noticed: look at the video's right-hand borders in VirtualDub. The right borders are warped toward the left at the top. That effect would point to a weak line tbc in the player, if the player has one. If you move the video frame by frame in Virtualdub, look at the sides of the back of the big chair and the textures and contours in the cloth; they change shape and ripple in almost every frame.....another sign of weak line tbc and unstable playback. Some of that is also caused by the way the camera recorded the scene. This adds to the noise problems and can cause artifacts in motion instead of plain old blur, which is easier to tolerate.

Avisynth has filters that can smooth but not eliminate most of these problems, and can clean up the grainy noise, the overall fuzz, most of the combing, chroma bleed (the blue shirt bleeds into the boy's arms), the bristling edges and other stuff. I can come up with some filtering ideas later. Several members will eventually tackle this and offer another 100 ideas on top of that. It'll take a little experimentation to decide which filters do the best cleanup without cleaning your video to death. What you do with those suggestions depends on how much you want to put into the project. I can say out the outset that Vegas, while it's good at what it does, won't help you much with most this work.

[VHS Roger]

Thanks a lot for the help. I have seen a lot of talk about Avisynth, but it sounded like it may be a bit over my head. Do you think I should still try it? At this point I'm hoping to just get the best video quality possible out of the captures I already have. It wouldn't be worth it to me to redo 50 hours of captures.

[LMotlow]

Here are two quickie trials. Both could be tweaked forever, but maybe an example of what can be done. Both are deinterlaced with QTGMC in AVisynth for double frame rate output. That will work for containers like mp4, but not for DVD. Playbac k errors are so extensive, an interlaced version hasn't solution hasn't come to me yet. The "MCTD" version used Avisynth's MCTemporalDenoise as the basic denoiser and cleaner. The "MCTD-NV" version is the same processing but adds a low-power NeatVideo in VirtualDub's output. Both are encoded using an x264 encoder. Black levels are a bit low, but I didn't address that.

I can furnish details later, but I don't think you can get these results in an NLE or in Virtualdub alone. Maybe someone has some ideas about what to do with all the scanlines that look like something out of phase.

[VHS Roger]

Wow, I think that looks amazing! I would like to be able to do that with AVisynth. I installed it, but really have no idea what to do. Would you suggest I look around for guides and try to figure it out or do you have any quick instructions? Thank you so much for your help!

[VHS Roger]

So I spent some time trying to learn how to use Avisynth. Here is the code I used:

avisource("original.avi")
ConvertToYV12(interlaced=true)
QTGMC( Preset="Slow" )
McTemporalDenoise(settings="high")

I'm sure I should be doing something different, but that was my first attempt. I'm not really sure how to actually save the file. I just used "Save As...". Would you mind posting the code you used and tell me how to properly save?

Here is what I made. Sorry it's so large, I don't really know how to properly save. It obviously has some issues.

[_Al_]

You open notepad, type that text/script, save it as something.avs , that is how you create avisynth script, it has avs extension.

Or , if avisynth is properly installed, you should create empty avs file right away, right clicking desktop and there you'll see an option to open empty avs file, same like txt file for example.

Then you load that script into a any x264 front encoder that can load avisynth script. Avisynth script behaves like video file (if avisynth is installed). As soon as you load it into aplication, avisynth starts to work in the background and serves video frames to that application as if it was video. That application could be VirtualDub, or encoder, or even video player like MPC-HC. So you can just test it, of course processing takes resources, so playback might not be smooth.

[poisondeathray]

I don't know if I should worry about going between YUV and RGB.

A potential issue with huffyuv in vegas, is it's not truly lossless with vegas - it's treated as RGB, so you will clip "superbrights" and "superdarks"

If you're planning to edit in vegas, you should ensure all the levels are "legalized" to Y= 16-235 if you are using a YUV intermediate, or convert YUV to RGB with a full range if using an RGB intermediate that you didn't previously correct for

(It's not just huffyuv - vegas does that with all commonly used YUV "lossless" codecs, like lagarith, utvideo, magicyuv etc..)

[VHS Roger]

You open notepad, type that text/script, save it as something.avs , that is how you create avisynth script, it has avs extension.

Or , if avisynth is properly installed, you should create empty avs file right away, right clicking desktop and there you'll see an option to open empty avs file, same like txt file for example.

Then you load that script into a any x264 front encoder that can load avisynth script. Avisynth script behaves like video file (if avisynth is installed). As soon as you load it into aplication, avisynth starts to work in the background and serves video frames to that application as if it was video. That application could be VirtualDub, or encoder, or even video player like MPC-HC. So you can just test it, of course processing takes resources, so playback might not be smooth.

I'm trying to use VirtualDub. I'm not sure if I add something to the code to save the file or use "Save As..." in VirtualDub. If I use "Save As..." the option to use a different codec is greyed out.

[_Al_]

Sorry, no, script is just a video, Avisynth always outputs uncompressed video, no information what format it should go into, nothing like that, that is up to software that you load that script into, in your case VirtualDub.
In virtualdub you can save it into avi container. What is in that avi container you set in Video/compression and choose appropriate codec (you perhaps aim for lossless, utvideo, Huffyuv, lagarith, they must be installed in your PC). You have to have set Video/Full processing mode as well. If you do not set compression, Save As would save uncompressed AVI.

[VHS Roger]

Sorry, no, script is just a video, Avisynth always outputs uncompressed video, no information what format it should go into, nothing like that, that is up to software that you load that script into, in your case VirtualDub.
In virtualdub you can save it into avi container. What is in that avi container you set in Video/compression and choose appropriate codec (you perhaps aim for lossless, utvideo, Huffyuv, lagarith, they must be installed in your PC). You have to have set Video/Full processing mode as well. If you do not set compression, Save As would save uncompressed AVI.

If I use Full Processing mode can I keep it YUV or will it automatically convert to RGB?

[manono]

You have to have set Video/Full processing mode as well.

Unless using VDub filters in addition to whatever's in the AviSynth script, you'll want to set Video to Fast Recompress.

[_Al_]

Yes, manono is right, Fast Recompress in VD, and then try this,
QTGMC works only in YV12 or YUY2 so Avisynth output is YUV, do not change colors to RGB.
You load that script into VirtualDub, export YUV, for example using utvideo you set YUV as well, fast recompress. I'd say this is better because you still have YUV lossless, as original.
Then load that into Vegas apply Sony levels effect, choose Computer RGB to Studio RGB, edit footage, export, it should keep original levels.

EDIT: I just tried with DV video, and Direct stream Copy from VirtualDub (utvideo, YV12 settings is ignored) Vegas handled ok, because it was uncompressed, but Fast Recompress needed that Sony levels effect Computer RGB to Studio RGB to keep levels, because it exported lossless. I correct that first sentence in this post then.

[VHS Roger]

I just tried with DV video, and Direct stream Copy from VirtualDub (utvideo, YV12 settings is ignored) Vegas handled ok, because it was uncompressed, but Fast Recompress needed that Sony levels effect Computer RGB to Studio RGB to keep levels, because it exported lossless.

I'm not sure if I want to use Vegas anymore after seeing how well LMotlow was able to edit the clip using Avisynth/Virtualdub. I'm wanting to splice together a bunch of smaller clips into one large video. I don't know if I should clean up the clips with Avisynth/Virtualdub, then put the movie together in Vegas, or just do it all in Avisynth/Virtualdub.

If LMotlow or anyone could recommend code to use in Avisynth I'd really appreciate it.

[_Al_]

Vegas is real videoeditor, it is more comfortable, you just need to remember to apply that Sony levels effect there after loading that lossless utvideo clip.

It seems to be more neat to apply filters in Avisynth, I cannot tell you what filters though, and use just VD to create lossless, whatever works. For terrible noise I just use neatvideo inside Vegas. If it is just joining of couple of clips do it in VD only. But you need to export it to some MP4 or something anyway. So you'd need another application anyway, like x264 front end or something to get MP4. Not sure what is your final delivery.

[_Al_]

For example, here is that what I did with your original, using QTGMC in Avisynth script, loaded into VirtualDub to get utvideo/lossless then loaded into Vegas using Neatvideo, exported to mp4. Just to get an idea what to get using neatvideo. You can choose to not filter that much etc. it is more user friendly then setting denoisers in Avisynth.

EDIT, that video was encoded wrongfully, I left --bff flag in encoder by mistake, it should be 59.940 progressive, but you get the picture ...

[VHS Roger]

Thanks for the help Al, that sample looks good. I'll need to learn to use Neatvideo. Did you use some kind of denoiser in Vegas/Neatvideo? I'm trying to use MCTD in Avisynth right now, but it seems very slow even for a small clip.

[LMotlow]

sorry to miss all the action, had to make a quick trip on family business and got stuck with a slowpokey Netbook.

I'd have to reconstruct from memory the script I used, since I used several versions and also played with MDegrain and other heavy hitters. For problem video I'd rather use small punches with a couple of denoisers together with one of the 'big guys', rather than shoot the whole thing to pieces ith one filter that won't do everything. I think others could come up with countless variations. Went something like this:

ConvertToYV12(interlaced=true)
AssumeTFF()
QTGMC(preset="fast")
MCTemporalDenoise(settings="high")
TemporalSoften(3,3,5,16,2)
ChromaShift(L=-2, U=2)
MergeChroma(awarpsharp2(depth=30))
GradFun2DBmod()
LSFmod(edgemode=2)
AddGrainC(1.5, 1.5)
return last

So many changes were made, I could be off on some values or sequence. For the NeatVideo version, or if you intend to apply VirtualDub filters or work in some app with RGB, change the last line from "return last" to

ConvertToRGB32(matrix="rec601",interlaced=false)  ## <- or = true, to suit the situation.
return last

For the NeatVideo version I used the same script but removed TemporalSoften.

I usually save VirtualDub output to Lagarith or UT Codec YV12. If I need something fancy like AfterEffects or further work, I save it in the required colorspace. I don 't care for using script-to-encoder in one step. Too many things happen and I change my mind a lot, so why run the whole slow process over and over?

NeatVideo: I don't use it often. It's almost always a final touch up, used with serious moderation. At full blast it turns video to Jello. I used it here to curb some of that "mushy" flutter on motion that you get often with tape. This sample ain't a fair test for the filter because I did't have a nosie smple area from the sample itself -- I used a noise sample from an old video with similar problems. The only way to use NV is thru its Advanced View setup. The more you know about Aviasynth and VDub, the more you can tweak NV's Advanced settings for best effect. Don 't use it's default settings. They're way too aggressive.

If I recall from looking at histograms of "original.avi", black levels are tad low. One could fix that in Avisynth with ColorYUV(off_y=3), or maybe off_y=5. This applies an offset that shifts all pixel values in the spectrum toward the right (brighter) side. A negative offset would shift everything to the left (dark) side. As it is, the brights in this sample won't be blown out with a slight shift. Sometimes, though, you have to get more complicated to keep your valid YUV and RGB levels.

Will try to keep up with the thread until I get home again. I hate Netbooks.

[VHS Roger]

I usually save VirtualDub output to Lagarith or UT Codec YV12. If I need something fancy like AfterEffects or further work, I save it in the required colorspace. I don 't care for using script-to-encoder in one step. Too many things happen and I change my mind a lot, so why run the whole slow process over and over?

Sorry for the newbie questions, but how do you save VirtualDub output to Lagarith or UT Codec YV12? The only thing I know how to do is use "Save As...".

I keep getting the error shown below. Any idea why? I haven't tried using any multi-threading with Avisynth. Not sure if that is a mistake...

[LMotlow]

If you're not using VirtualDub filters, but basically using VirtualDub to monitor Avisynth output and save the results to a new output file:
- "Video..." -> "color depth....." -> set the colorspace you want for output.
- "Video..." -> "compression..." -> set the compression you want for output.
- "Video..." -> "fast recompress"
- "File..." -> "Save as AVI...." -> set the location and name of your output file.

If you're not using VirtualDub filters, but basically using VirtualDub to send Avisynth output to a new file using the same colorspace and compressor:
- "Video..." -> "direct stream copy"
- "File..." -> "Save as AVI...." -> set the location and name of your output file.

If you're using VirtualDub filters:
- "Video..." -> "color depth....." -> set the colorspace you want for output.
- "Video..." -> "compression..." -> set the compression you want for output.
- "Video..." -> "full processing mode"
- "File..." -> "Save as AVI...." -> set the location and name of your output file.

I finally found the copy of the Avisynth script I used for the MCTD videos posted in #9 earlier:

ColorYUV(off_y=3)
ConvertToYV12(interlaced=true)
AssumeTFF().QTGMC(preset="fast")
Vinverse()
MCTemporalDenoise(settings="high")
TemporalSoften(4,4,8,15,2)
ChromaShift(L=-2,U=2)
MergeChroma(awarpsharp2(depth=30))
LSFMod(strength=75,edgemode=2)
Crop(0,0,-12,-8).AddBorders(6,4,6,4)
return last

[VHS Roger]

Awesome, thanks so much for the help, I'll try that out.

One question: is there any reason to use crop/addborders like you did versus using something like BilinearResize?

[KarMa]

Just putting this out there, QTGMC spatial denoises by itself automatically. Other people on this site have talked about using QTGMC just for its denoiser. So keep that in mind, you are not just temporal denoising. I'm not saying that is a good or bad thing.

edit: here is the thread https://forum.videohelp.com/threads/358043-A-comparison-of-AVIsynth-denoise-filters

[LMotlow]

Awesome, thanks so much for the help, I'll try that out.

One question: is there any reason to use crop/addborders like you did versus using something like BilinearResize?

Severalreasons. 1) cropping usually is uneven, more pixels from one dimension than the other. Resizing will alter the original image aspect ratio. 2) BilinarResize will cause sawtooth edges unless it's 1:2 resize or per-pizel resampling,. For soft video like this sample, BicubicResize will make it softer. There are better resizers in Avisynth. 3) If the video was sharp and pristine, resizing wouldn't hurt, but resizing always has a quality cost more or less, usually more. 4) The borders were cropped and replaced with new borders to center the image and maintain the original frame size without changing the original image proportions.

If played on TV, most new TV's have overscan enabled, which hides thin black borders. Even if you disable overscan, 4 to 6 pixels on each side won't make much difference. Many HD TV shows and movies have borders of one kind or another, for this reason.

A lot of that denoising was more or less standard stuff, but I used more than one for different reasons. Some of the motion in the original looks like motion interpolation errors. QTGMC and MCTD together can smooth out some of that stuff. The boy's hand and the girls head motion and pigtail look really odd in the original every time they move.

[Brad]

Just putting this out there, QTGMC spatial denoises by itself automatically. Other people on this site have talked about using QTGMC just for its denoiser. So keep that in mind, you are not just temporal denoising.

I thought QTGMC's denoising effect was caused by its motion compensation, thus temporal...?

[LMotlow]

It is. But it's not enough here, and there's almost no strong edge or detail delineation in the video to play with. So I used a weaker QTGMC, if for nothing else than to get it deinterlaced, and MCTD for the other stuff. I did use QTGMC by itself with loads of parameter changes, along with playing with other denoisers, but it looked like either too much or too little and there was still bad pixelation on moving details. A strong QTGMC didn't leave much to work with, there was still a lot of grainy noise, and the motion artifacts persisted. Dfttest and MDegrain (QTGMC uses both) by themselves had the same problem. I'm sure there are a load of other ways to work that clip.

Any ideas where pixelation comes from here? Look at the boy's hands.