The sample you posted (test3.avi) is uncompressed YUY2. You should be using a lossless compressor.

10% dropped frames is crazy high. It's possible your hard drive isn't keeping up with your uncompressed capture, in which case using Huffyuv etc. could solve this. Even uncompressed SD isn't very taxing for modern HDDs, though.

Thanks msgohen - I installed the lossless compressor from this forum but not sure I activated it. How do I do this in Vdub?

Go to Video -> Compression while in the Capture mode.

Thanks - I have enabled Huffyuv v2.1.1 in the compression settings and whilst it has significantly reduced a 16 minute file to 8.3Gb, it has introduced some other problems; namely audio sync issues and a slow speed - I would attach it here but I need to know how to edit a 16 minute capture to reduce the file size to less than 99Mb to enable me to upload ?

Also, I am still dropping around 13% of frames so out of 22461 frames, I dropped 2930 - does it matter if the drive I am capturing to is not the OS drive but another SATA drive connected to the same motherboard ? Just going back to the previous issue with audio sync and slow speed, could it be because 2500 frames were inserted ?

Looking around at my settings, does it matter if the "Capture PIN" menu properties show YUY2 but in the "compression" menu, Huffyuv is enabled? Do they both need to be the same ?

Hmm. Not sure what "slow speed" means.

Capturing to a secondary drive is the recommended setup and is much preferred over capturing to the OS drive, so it sounds like you are doing this properly.

If I'm to understand that "slow speed" means motion is jerky, dropped and inserted frames will both cause that. The inserted frames are actually VirtualDub's attempt to keep the audio in sync by adding frames.

That is the correct setup. The hardware outputs uncompressed YUY2, which is compressed to lossless via software.

Try a Test video capture (F7) instead of a real capture, during a portion of a tape that contains only normal signals: no starts/stops/blank unrecorded sections. This will capture the video & audio but discard it rather than writing it to disk, to help troubleshooting. Do the stats still show the same number of drops and inserts with Huffyuv? What about with uncompressed?

We won't need 16 minutes to spot problems. A few seconds of lossless huffyuv YUY2 would do it, taken from the end of the capture where the audio is definitely out of sync by a noticeable amount.

You can cut out short edits from longer captures and create samples without recompressing or reprocessing the edit, by changing "full processing mode" to "direct stream copy" in VirtualDub before saving your edited sample. Those mode settings are in VDub's "Video" top menu.

Your test3.avi sample was a 3-second uncompressed YUY2 video in 75MB. If it had been lossslessly compressed with huffyuv or Lagarith, the same 3-second file would be 29MB. Typically, YUY2 compressed with huffyuv will fit about 10 seconds of PAL 720x576 YUY2 video into 90MB.

You have bad audio sync and unstable frame speed and dropped frames because you need a frame-level TBC. You risk slower disk i/o and more dropped frames by capturing to the OS drive.

By default, VirtualDub captures to very large uncompressed RGB unless you specify (a) a different colorspace, such as YUY2, and (b) a lossless compressor such as Huffyuv or Lagarith, either of which can compress your YUY2 capture losslessly. If you specify YUY2 but do not specify a compressor, your capture will be uncompressed YUY2 and will be about 2.5X to 3X the size of the same video losslessly compressed with huffyuv. A YUY2 video after lossless compression is still YUY2.

Thanks both - here is a small clip :)

Attachment 6439

I should clarify what I mean by slow speed - by this I mean Slow Motion i.e. the video is running at what seems to be <1x speed but still smooth and no jerkiness and the audio is out of sync but also has slow motion speed.

Also, stats are bad in uncompressed and compressed mode

I should also mention that it starts fine but 15-20 seconds in and all the problems begin

For the Direct Stream Copy sample:

Assign the start frame of your sample by hitting HOME on the keyboard.
Assign the end frame by hitting END.
F7 to save your sample clip.

Thanks - uploaded to post #108 :)

The low-pitched voice is caused by VirtualDub's resampling function trying to correct the sync and badly over-correcting.

See whether sanlyn's Timing Options are any better.*

*EDIT: Umm. The frame rate of the clip is 30.00 fps. Test3.avi was 29.970. You need to set 25 fps in the capture options for PAL. Either Capture Pin or Capture -> Settings. Don't touch the Timing Options until you do another test capture and see whether the correct frame rate is enough to fix things.

The motion is not smooth at all. Many duplicate frames.

Yes that is MUCH better thanks - it even helps with the dropped frames, now only dropped 6 frames out of 21198 - do I even need a TBC with this low number of dropped frames ?

See below for the capture - any further advice or am I good to go now ?

Hanover bars are visible, but I don't think you can do anything about this on the capture side. You would deal with them as part of restoration filtering.

Thanks msgohen - how do I process hanover bars with VDub ? Where can I find suitable filters for this noise ? Other than that does it look ok to you ? Do I need a TBC if it only dropped 6 frames out of 21198 ? Should I wait until I have a TBC before I start mass capture ?

I'm not aware of a VirtualDub filter for that. I never tried it with VirtualDub, only with Avisynth. Any such filter will soften somewhat, more or less. Others might have a suggestion.

Black levels are too high. Color and luma values are compressed into a very small area of the spectrum and the video looks fogged out. For some reason, facial highlights and contours look as if they're being posterized. Does that video have built-in DNR, or did you apply any denoiser during capture??

Yes. Why would want to purposely insure dropped frames? That rate of 6 drops for 14 minutes, if it continued at that rate, would throw audio out of sync by about 1 second after an hour. If you want, you could get by on the cheap by using a used DVd player such as a Panasonic Es10 o ES15 for pass-thru with decent frame sync (it won't defeat Macrovision, though, if that's a concern).

I recorded a couple of VHS tapes off cable that had so much signal noise, I was getting Macrovision effects and image distortion during capture. Cured only by using a full-frame TBC.

Thanks Sanlyn -

Are there any Avisynth tutorials for hanover bars ? I have done a few searches but it seems that you need to be clued up in scripting, is that right ?

To answer your question on DNR, yes there is something called "Intelligent Picture" on this machine which I can disable or set to manual where I adjust the sharpness myself. Shall I disable this permanently ? How do I change the black levels and Colour/Luma values ? Do I do this at capture or post capture ?

Thanks for your advice on TBC - I tried a full 3 hour tape and while the frames dropped were around 25, the audio still seemed to be in sync so it wasn't noticeable but I will try to secure a TBC - they seem to be quite rare especially the DataVideo ones.

Other than that, is there any other advice that I should take note of before I start capturing en masse :)

That's right. Hanover bars usually affect chroma, not luma. Cleanup is best done in a YUV colorspace, which stores chroma data separately from luma. In this case the YUV matrix would be YUY2 or YV12. Low-pass and other filtering techniques have been used, requiring Avisynth.

Frankly I don't think the Hanover effects are enough here to lose sleep about. A more visible problem is false contouring and rough edges instead of smoother gradients or transitions (looks like overly aggressive degraining). Highlights seem a little hot, even when they're not bright. Some annoyances: open your sample "David Bowie interview 25fps sample.avi" in VirtualDub, advance until you get to frame 173, and look at the top of the frame. Besides the thick cyan bar at the top border, all other frames have a line of yellow "dashes" across the top. Quick deinterlace with yadif in VirtualDub shows that these dashes appear only on even-numbered fields. The same pattern is in your other samples.

25 PAL frames is 1 scond of video -- not disastrous, but not a good sign either. If you were dropping just as many audio frames, which normally happens, you'd be out of sync. You might have missed the comment about tbc pass-through alternatives near the bottom of post #115.

DNR and sharpening are different things. You can try sharpening less, as VCR sharpeners also sharpen noise. Disable Intelligent Picture for a short capture and see what you get. Post processing is more advanced than most VCR sharpeners or DNR.

How did you make the dark "test3.avi" in post #95 to be so much brighter when you captured "David Bowie interview 25fps sample.avi" in post #112?

More importantly If you have clipped darks or brights, it's too late to repair it later. Valid RGB 16-235 levels are checked before capture with the VirtualDub capture histogram. This is especially important since you appear to be using an uncalibrated monitor. Using the histogram are covered in several recent threads. Here's one of them: Using a capture histogram to check proper levels.

There's a learning to curve to this VHS capture and restoration stuff. Capturing en masse will be very overpowering. I'd start with one or two caps and use them to learn the ropes. Later it gets easier and faster.

Sorry for the radio silence again - been trying out different things recently and have come to the conclusion that there is a fault with my VCR as it has colour bleed/smearing issues. See attached clip for what I mean (also sorry for the general poor quality of UK TV - I can't be responsible for this output :))

If you look on the lady's multicoloured top, you will see that there are black horizontal lines going onto the dark part of the jumper...can anyone confirm if this can be cleaned up or whether there is indeed a problem with the video player ? I should point out that the tape itself is nearly 20 years old so that may be part of the problem here

Also on a second clip I will attach to another post, there are black flashes at the top of the screen - again would appreciate any advice on this i.e. whether it is hardware related or the age of the tape (similar age to the first clip)....(note if you cannot see the flashes at the top of the screen, please adjust your brightness on your monitor - it was only when I viewed this clip on a full TV that I saw the flashes of bright light)

By the way I have also been playing around with the histogram feature and there is a big difference between what my eyes thinks are good images compared to what is actually a good RGB spread on the histogram !

Thanks for all your help

Sorry for the delay. Busy with health matters of elderly relative.

Part of the dark lines is in the jumper itself, but the evenly spaced bright/dark pattern is ghosting from the bright contrasting cloth on the arm. This is common with older cable broadcasts and was likely broadcast that way. No effective fix without making things look worse. Also caused by smearing from dnr circuits in many VCR's, but this looks like a defect in the original transmission.

The flicker across the top of the frames is in both samples. On a properly calibrated monitor you don't have to increase brightness to see it. It looks like mistracking or improper tape alignment. Your VCR likely has a tracking adjustment control that could mitigate it. The original tape might have been misaligned during original recording. If you view the brighter flicker frame by frame in VirtualDub you'll see that shapes of objects subtly shift position and shift back again. I don't see much color bleed, but an Avisynth technique used in the original YUV colorspace can make it barely visible.

The capture histogram is a valuable tool, glad to see you're getting the hang of it.

The samples are appreciated, but if you're concerned about file size you should know that both samples are uncompressed RGB, which increases file size. I compressed your samples to what I hope was the original lossless huff/YUY2 and get the following results:

test colour bleed edit.avi: uncompressed RGB = 75.8MB, huffyuv/YUY2 = 41.9MB
white flashes.avi: uncompressed RGB = 99/MB, huffyuv/YUY2 = 37.3MB

Hi Sanlyn,

Apologies for the long silence, I have had other priorities recently so this has taken a bit of a back seat for now.

Thanks for your input, have taken note but please could you elaborate on the Avisynth technique you refer to in your post for further information ?

Also, now I have had a chance to make a few captures, I want to know how best to copy to convert from an avi file into an mpg or other format? I have seen a guide on this very useful forum to download something called Avidemux and configure in the audio and video formats (here: http://www.digitalfaq.com/guides/video/encode-mpeg2-avidemux-pt2.htm) but I have no idea how to tell whether my source is interlaced or not and furthermore what the optimised settings are given my input source. Furthermore, there is a tool called Gspot which identifies whether your source is interlaced or not but when I run it, it doesn't tell me whether my sources are interlaced or not.

Also, on this thread here: http://www.digitalfaq.com/forum/video-editing/6848-merging-avi-dvds.html both you and Lordsmurf refer to the steps to convert to a DVD but I have no DVD authoring software - could you point me in the right direction for this and also what is meant by "MPEG-2 to DVD-Video compliant specs" mentioned in post #5 by Lordsmurf.

Finally, when converting the AVI files into MPG format, is there a quicker way of doing this as it seems to take a very long time to convert i.e. it seems to play it back one second at a time to do the conversion which could take several hours if it is a very long file. Is there any other way? Also, as part of this question, how could I shorten the video down beforehand to make the conversion specific to the start time and stop time I want to capture from and to ? For example, if I have a 20 minute clip but I only wish to convert from 10:00 to 12:05 (down to specific frame number) then how should I go about shortening the file first then converting or do I convert the entire 20 minute file then shorten it ?

Thank you for all your help in getting me this far.

P.S as an unrelated question, I have some MPEG.TS files that I want to edit (i.e. remove adverts from) - which software should I be using and what are the steps?

Suggested workflow:

1 - Make a nominally workable lossless capture, attaining proper video levels and tuning player settings for optimal playback. You are still in this stage and learning.

2 - Lossless Cleanup, denoising and repair of common analog defects, color correction, chroma cleanup, saving the results as lossless intermediate working files. Tools: Avisynth and VirtualDub. Along the way, use Avisynth and VirtualDub edit features to discard material you don't expect to keep in the final version.

3 - Lossless input into timeline and/or encoding editors for joining, transitions, special effects and further edits using lossless media, and outputting encoded final delivery formats (DVD, BluRay. AVCHD, MPEG, h.264, etc.) and eventual burning to disc or web posting. Possible choices: Corel (formerly SONY) Movie Studio Platinum, Premiere Elements, or TMPGenc Video Mastering Works, all of which can join losssless input, apply various effects such as transitions, can author for burning, and can burn to disc. You can use freebies like AviDemux if you want but it's basically an encoder -- there are dozens of such freebies, but all of them are more limited than a good NLE package.

4 - For burning to DVD or BluRay disc, most NLE packages that can encode to those formats will also burn to disc. The exceptions are some freebies that are encoders but not authoring or burning apps. An excellent free disc burning app is Imgburn, which some freebies implement internally.

Capture, restoration/repair/correction, joining, applying effects, encoding, and burning to disc are all separate steps. There is no single package that can do it all, even if they advertise that they can. In particular, editors are a very poor choice for restoration and cleanup.

I'm not sure what you mean by optimized settings based on source structure. Analog tape is captured straight through from player to capture device to capture file. Optimal capture settings are directed at valid and workable luma and chroma input levels to avoid clipping brights or crushing darks. If you're thinking of doing something wild such as deintelacing on the fly during capture, all bets are off -- few users in any forum would be willing to help fix video that's purposely damaged during capture. PAL and NTSC analog tapes are almost always interlaced or telecined, and so are tapes made from TV shows. Tapes made from movies are usually telecined, tapes from home movie cameras are interlaced. Some PAL movies are telecined, some use pure interlace, some are speeded up from 24fps to 25fps for PAL.

GSpot hasn't been updated in a decade, even if you can get info such as GOP size and whatnot from pre-encoded material. Use MediaInfoXP (http://www.afterdawn.com/software/au...ediainfoxp.cfm or http://muldersoft.com/#mixp). Don't depend 100% on media utilities for the full story. Specifically, many telecined films are listed as "progressive" and often as "interlaced", when in fact many could be progressive with hard-coded pulldown. Learn to determine these things for yourself: Neuron2_How To Analyze Video Frame Structure.zip .

You'll either be embarrassed or amazed by the answers to those questions. No one would encode an entire capture to a lossy format and then start chopping away at it, unless they have no understanding of what they're doing. Capture your tape to a lossless file. Then cut the losssless capture as desired with VirtualDub or avisynth. Then make fixes or corrections to your lossless cuts. Then join the pieces in an editor, and encode as the last step.

If it looks as if your encoder is encoding twice, it isn't. It just looks that way, maybe because you don't understand two-pass encoding, which is preferred for most standard formats. The first pass determines the best way to utilize the bitrate you've set up. The second pass is the actual encode. Best for efficient use of bitrate, best for cleanest output and optimal file size. Encoders are slow because they're encoders, not players or simple copying apps.

Then by all means, get some. TMPGenc Authoring Works, Movie Studio Platinum, Premiere Elements, can all work with lossless media input and can output to multiple formats. The one to avoid is cyberlink. You can go for freebies like DVDStyler, AVStoDVD, DVDauthorGUI, etc., but be warned that freebies are generally more difficult to learn, have format limitations, and have fewer features than paid software.

common PAL specs for valid DVD encoding:

DVD disc folder and file structure are built and formatted by authoring programs for DVD authoring/burning:
http://www.videohelp.com/dvd#struct .

Avisynth corrections are used with lossless capture in their original captured colorspace after capture, and not used on encoded video. For what it's worth, some of the techniques used to correct levels, chroma bleed and color shift, and defects such as dot crawl or occasional Hanover bars are: the ColorYUV and SmoothAdjust filter for lume and chroma levels, low-pass filters (targeted resizing) and masking of Chroma channels with filters such as FixChromaBleeding() and awarpsharp2(), and smoothing/deblocking filters such as GradFun2DBmod or Gradfun3 for handling posterizing and hard gradient affects. The latter effects are due to overly strong denoising during capture. To get an idea of how these are used we would need some lossless samples that exhibit those problems.

At this point I'd say you're still learning to capture for optimal results that can save a lot of work later. It's good to plan ahead, but don't jump the gun by getting into a rush with troublesome captures.
:wink2: