VHS capture loses sync with home movies, not VCR tapes?
 

I am in the process of capturing old home movies taped on a camcorder as well as some tapes my dad taped off tv in the late 80s (tv shows, etc). I find that capturing these shows and editing them later results in no A/V desync problems but editing the segments with home movies will ALWAYS give me a desync the longer the video plays out. (by the end of a 1hr segment the audio maybe off by 2 seconds). Is there something Im missing during the capture phase?
(The camcorder was a GE CG-9908 with fixed SP recording mode)

Here is the workflow:
Sony SLV-575UC vcr > Panasonic DMR-ES15 dvd passthru > Hauppage USBLive 2 > Hauppage Capture software (usually 5000kbps constant btr)... and then Vegas 15 for editing and resaving

I am aware of AmaracTV however these can result in 50GB+ files for 1 hour recording which makes it difficult to do batch captures and editing later.

I can post Mediainfo screenshots with the initial capture file vs the edited saved file that desyncs if needed. I tried to match every setting down to interlaced first frame, GOP, and the other mpeg2 settings in Vegas

I do not understand if the sync is lost at capture level or later when editing the capture with the NLE...

What form/format are the captured files? Is the amount of de-sync the same for all tapes, or are some camcorder tapes worse than others.

At one time video and audio could get out of sync due to slight frequency differences in the clocks used for the video and audio in the capture and editing process.

Also I've seen slight de-sync in AVI files driven by occasional dropped frames during capture. That it happens with only camcorder tapes implies issues with those tapes. One possibility is if the the camcorder tape contains glitches (e.g., at stop/start points in the recordings made without flying erase head) that glitch could induce the de-sync in a continuous capture.

If the tapes contain glitches is the sync steady between them? If so, using your NLE you can correct by making cuts at glitches and then offsetting the audio slightly for each clip.

You are correct, the unedited capture file plays fine, however my dad often used tapes with other footage on it to make our home videos so editing is important as there are dropouts, segments where the previous recording shows through, etc that I would like to clean up

Sorry I am new to the site, maybe it would've been better in the editing category, but I was concerned I am not capturing it correctly. Perhaps this should be moved then?

Perhaps that is it, Most of these home movies have stop/start segments (opposed to one continuous 20min recording). I could try to go minute by minute and as soon as I notice desync and move the audio to match. Like I said above, these captures do need editing otherwise I would leave it as the untouched capture (which does not have sync issue) and call it a day.

The hauppage capture software records as either as .ts or .m2ts with a choice between a fixed/variable btr up to 10,000kbps.
Here is a screenshot of the capture information for an untouched file.

Attachment 14486

I'm sorry to say this but your capture workflow makes no sense.

Why not just record with the ES15 in mpg at the highest possible bitrate (1 hour ~ 4,3 gb)?

Transfer the whole thing to the computer with DVD's or Iso Buster and then there are enough programs with which you can cut the mpg files without losing syncronity. A re-encoding is not necessary if you cut only on the I-frames (for example I have used the freeware cuttermaran but there is other freeware available). If you wish to cut frame accurate a tool that supports smart rendering where only the cutting area is re-encoded is what you need.

Then use the free software hybrid+vapoursynth filter to improve your recordings and encode to mp4/mkv.

If you want to use the Hauppauge then only to record lossless. There are no limits for editing and post-processing your files. Without any sync problems.

Is 50GB really a lot for one hour these days? A 4 hour recording should be about 320 GB. Cutting the whole recording in Virtual Dub is done in a few minutes if you have done it a few times.

I can only recommend you to think about your capture workflow and the image quality.

Thanks for your reply, now is clear.
It appears to me that the problem is in your NLE, so you should first of all degug there what happens to fix it.

In addition something you can try:

• capture smaller segments
• the captured mpeg2 transport stream relies on internal PTS to synchronize audio and video packets; it is worth to run a small utility that was used 30 years ago to check and eventually adjust the time stamps if there are errors inside the stream because hauppauge software does not report dropped or inserted frames (PVAStrumento, attached). Better softwares exist todays, such as TSDoctor or VideoRedo if you wish.
• provide to your NLE demuxed (and synchronized) audio and video streams (use options in PVAStrumento) instead of full mpeg2 ts.
• the real time mpeg2 capture of the Hauppauge software + hardware is not bad at 10 Mbps (and btw is better than a mpeg capture performed by the ES-15). Here a comparison with a HuffYUV capture (compressed to mp4 for size reason, but with a crf at 17 to impact as small as possible): Attachment 14489
  However, as Bogilein suggested, a lossless capture is better, especially if you need to further edit the videos. A cut without decode/encode is only possible at the end of a GOP, for example, and any additional decompression/compression reduces the quality. Providing your NLE with a lossless capture is better compared to mpeg2 stream for quality but also for the manipulation of audio/video: in the avi file the synch of the audio and video streams is similar to pts (interleaved with audio timestamps) but easier by nature.

I have been told that nowadays computer capture is better than what the DVD capture then would be capable of, but as for the second part of your suggestion, I will keep that in mind

Thank you for that software. I have tried it out and when scanning one of my .ts files from the home movies it indeed found missing frames and GOP errors. However after running through the demuxer with error correction, it created a new separate audio and video file with uneven lengths, and I have tried to line up the audio to the head, tail and somewhere in the middle of the video stream with no sync success. Is there something I'm missing?

All else fails I will fall back to use amarecTV for the home movies. Previously i had been using the Lagarath (lossless) codec which produces an AVI file with a YUV422 color sample, is this satisfactory?

Can you post the PVAStrumento report?

If you have timing structure errors, PVAS should be ablo to fix them, and provid either a synchronous mpeg2-ps (program stream) or a set of demuxed elementary streams synchronous (if there are not tons of them in a short timeframe).

BTW, the fact that you have errors in the capture is not a good sign if they are in the middle of a proper video sequence.

EDIT:
Yes.

I've attached the full report:
Attachment 14501

And a sample of what it contains:

ALERT> 2 video frames missing at 00:04:24.532.
ALERT> Dropping GOP for sync.
Re-syncing.
Closing starting GOP.
Aiming for sync at 00:04:27.188
Cutting video.
Closing starting GOP.
Aiming for sync at 00:04:27.402
Cutting audio.
Audio stream 1 now starts at 00:04:27.403
Video stream 1 now starts at 00:04:27.402
ALERT> -1 video frames missing at 00:04:27.402.
ALERT> Dropping GOP for sync.
Re-syncing.
Closing starting GOP.
Aiming for sync at 00:04:27.802
Cutting audio.
Audio stream 1 now starts at 00:04:27.811
Video stream 1 now starts at 00:04:27.802
ALERT> -1 video frames missing at 00:04:27.802.
ALERT> Dropping GOP for sync.
Re-syncing.
Closing starting GOP.
Aiming for sync at 00:04:28.202
Cutting audio.
Audio stream 1 now starts at 00:04:28.195
Video stream 1 now starts at 00:04:28.202

Now to demux, I press the demux button, and let PVA make a separate audio and video file, which should be synchronized when done? here are my settings:
Attachment 14502

is probably a false postitive. Try to set "Maximum PST-Jitter" to a valus < 33367 microseconds in "Expert" tab (1/29,97 frame per seconds for NTSC)

The demuxed streams are normally synchronous, whatever their duration is, and can be later used as such.

In alternative you can use "make ps" to work with a single mpeg2 program stream file, where the audio/video streams inside should be synchronous, and you can try to import it in your NLE.

BTW, a real time compression is resource consuming, so while capturing do nothing, disable all processes not related to capturing, stop antivirus, disable wifi or internet connection, etc... Do not even move the mouse ;)

Thanks for being patient with me :) Currently the max jitter in expert was 5000microseconds, so then I set it to 33367 (?)
Even when demuxing again, the files come out with a 10 second length difference.
Attachment 14506
And then I tried to make a ps file only for it to also be unsynced! Really confused here.
Soon I might have to babysit the recording and have my finger on the pause button so I can get only the home movie videos i want without needing to edit anything :unsure: haha

Load somewhere your capture and post or send me the link

Thank you for all your help so far. If this issue can be resolved then I can do much better for other 15 tapes or so that will undoubtedly have a similar issue
https://www.mediafire.com/file/qeqtu...lollo2.ts/file

The video file is there as a .ts file
Also I just noticed you are in Italy, I spent a few months there on a school trip about 10 years ago, did a tour from Ascoli Piceno down to Matera and back up to Rome and Florence. Wonderful country :)

PVA Strumento has definetely a problem with your file. I used TSDoctor to generate a new clean transport stream file and then to demux to elementary streams. The old .ts file, the new .ts file, and the 2 elementary streams are synch in my experiements (even if in your video is not easy to check for a lip-sync at the end).

If you use the new files in your NLE it should work (but I guess the original should work as well).

I sent you a PM with the links to the files ;)

Yes I realise these movies dont have much to go on in terms of matching mouth and audio visually :p I looked at both the new .ts file and the demuxed files and through the NLE seemed to do a better job with it :) B
I downloaded the trial of TSDoctor, and ran the original video through it and curious enough it didnt detect any issues with it, was it the same with you? I ran a few of my .ts home movie capture files through it and didnt have any issues with scanning the file at least

The Panasonic recording quality is terrible.

I'm assuming Vegas is capturing lossless/uncompressed, and then exporting MPEG. Not actual capturing direct to MPEG.

Never use NLEs for capture. Too much overhead. Many lack dropped frames counters, others just inaccurate. We don't use the freeware VirtualDub (or AmarecTV, for some cards in Win10) because we're poor (ie, cannot afford Vegas). It's the best tool for the task.

Why are you getting TS files?

Remember that the ES10/15 is not a TBC. It does not replace a TBC. It's a mere strong+crippled line TBC with non-TBC frame sync, and it really has no frame corrections beyond bare minimums.

I agree with all of that. I am not using an NLE for capture, only for later editing, as I said earlier, these tapes have more than one program which is unwanted in the final file so is easier for me to record the entire tape in batch than to sit and watch every minute with my finger on the pause button during real-time capturing.

The reason for me not getting an TBC is simply, cost. Nobody in my family has a urgent desire to see our home movies in lossless archival quality, and I cannot justify spending $1000 to capture 10 hours of home movies that may be watched once a year if that... Now if there was a way to RENT a tbc for a day or two..sign me up :D

The "Hauppauge Capture" software generates by default an mpeg2 transport stream. Used in broadcast contest, a transport stream contains more additional data, to synch audio/video compared to standard mpeg2 program stream. I suspect that choice was made by Hauppage because the real time capture is sensitive to problems

I do not think OP is doing that, just importing the capture in a NLE. He had problems with this operation.

Yes, that's why I said PVAS had problems. TSDoctor is a wonderful piece of software, I use it every day for my DVB-S and DVB-T captures (dumps is a more appropriate name)

Edit: OP replied before me ;)

Let me ask this question. I have a thing called a line doubler, basically meant for SD signals from old video game consoles to be able to be displayed on modern TVs. I have read some of these can actually be used in lieu of a TBC because it improves the image by its own nature. Here's a description of the device as used for VHS/LD capture:

"The RetroTINK-2X can be used as a video capture device but is not an optimal solution since it only incorporates a 2-D comb filter and has low tolerance to VHS errors since there is no frame-buffer for low lag operation. LaserDisc video output by a player with a built-in comb filter via S-video should work much better.
You can either output video line doubled (bob de-interlaced) to 480/576p which should work with most capture cards. It is also possible to output interlaced video but compatibility will be much poorer. We have verified interlace capture with the DataPath Vision E1 and OBS"

So basically it can output 480i and 480p via HDMI.

Now I have the E15 DVD player which i believe has a built in comb filter. Would it be worth a try to get an HDMI capture device to feed the 480i signal right into my computer/Vdub?

When I watch VHS on my tv (ie not recording anything) the workflow is:
Sony VCR > E15 DVD passthru > linedoubler/upscaler via svideo input > HDMI to OLED TV
I am able to FFD/REW without picture dropouts (unlike without the dvd player where anytime the picture moved too fast the signal would drop)

Line doubler? You're making things worse for yourself.

RetroTINK is for video games. Not VHS. I don't care what anybody says or writes, the proof is in the bad capture output.

Between it, and the Panasonic DVD recorder -- both lousy choices -- I'd opt for the DVD recorder.

Ok thanks just thought Id ask since Vdub captures at 480i anyways from my USB analog capture device

Gave it a lot of thought and went back and started recapturing the home movie tapes from scratch via lossless
Tried Vdub first but always ran into heavy sync issues (there would be dozens of frame drops in the first 10 seconds).
Used AmarecTV instead with LAGS. File appeared to be OK... but there was a moment when the picture froze and the audio kept going, but still managed to be in sync.

The end plan for these videos are going to be computer/file viewing, no real intention for DVD or BD. But any time I process the video stream, that freeze spot creates an audio desync for the rest of the video length. Last night I tried Avisynth and QTGMC for the first time... looked beautiful, smooth and deinterlaced, but the audio was mismatched at that spot in the new file.

Is there anything I could differently during initial capture to prevent glitch spots like this? Or is this a TBC only fix. I attached the segment of the video when it visibly freezes then jumps ahead.
Attachment 14564

All that aside, now I really understand the errors of my ways when comparing the 5Mbps m2ts capture to the lossless one I just made last night, all that ugly color noise gone.
Attachment 14566 Attachment 14567

I'll add this this post is getting me closer to the issue, but AVISource gives me an error that it cannot find a decompressor for LAGS.
http://www.digitalfaq.com/forum/vide...gmc-audio.html
https://forum.doom9.org/showthread.p...27037&page=130
So if it still a TBC issue I'll seek an answer here, but it if is truly just an issue with processing improperly I'll keep searching around.

I am calculating 364 dropped frames after deinterlacing.

Welcome to the club of lossless captures!
However, at 10mbps I did not see major degradation between lossless and mpeg2 capture. The "ugly color" difference may be related to mpeg2 compression (difficult to believe), colorspace conversion (YUY2 -> YV12) for mpeg2 generations, capture setting or who knows what. In any case, stay with lossless captures!

Concerning the delay/freeze problem, avoid to open lossless files with "mpeg/avc-built dlls", such as Ffms2. Capture with HuffYUV and open the file in AviSynth with AviSource("<filename.avi>")

Thanks that helped a lot. As for the color blocking it is hard to say. Happauge captures a m2ts or ts stream at 4:2:0 and thats pretty much the end of it. LAGS and others capture at 4:2:2 which may have something to do with it.

I shouldve done this a long time ago but I didn't. I may have stated at some point I have two VCRs I use during this project, one is a 1990 Sony SLV-575UF, decent unit with a pull down station and swing shuttle. The other is a 2001 Panasonic PV-V4211. Both units are Hi-Fi 4head. Typically I use the pana as a watching unit (seeking through tapes to write down record times) and use the sony to actually capture to PC since I assumed its a better higher end unit..

I captured the same tape on both units and was shocked to see the Sony looks WORSE! Is there a cause for this? It has a sharpening knob but its always in the middle. Both files were the same codec lossless.
FIRST: 2001 Panasonic
Attachment 14585
SECOND: 1990 Sony
Attachment 14586

As per today, the working conditions of the machines are more important than the initial specifications/features. Try both on the field with several segments and choose the best