Help needed - how to strip, edit, and rejoin audio/video

[Coop]

I captured an older VHS music video compilation twice over the past two nights. Using Virtual VCR, the first capture had a few dropped frames, the video bitrate started low - 29.94 and slowly crept up to 29.96 fps, and the audio was scratchy and in some cases distorted. This led me to believe my VCR need a cleaning (I've had it for 4 years, this is probably the most use it has ever had). I cleaned it, and the second capture was better, however the audio is still a bit scratchy.

The capture is in AVI format. Does anyone have a "how to" on stripping the audio out, what to use to edit, and then how to blend the video back together with the audio? I'll be burning 720 x 480 DVD, so the final format will need to be MPEG2.

Thanks......

[SaSi]

The first thing you must get right is capture. Having dropped frames will cause a messy situation of audio/video out of sync and you DON'T want to mess with solving such problems.

I have had severe problems trying to capture from VHS VCR and tried several programs (including fancy ones that came with the ASUS 9180 Video Card).

The problems were solved by means of VirtualDUB and DIVx. Things should work nicely if you have a P4 class machine with a CPU or a P3 class machine with a CPU better than 1,000 MHz.

1. Prepare a Hard disk for capture. If you have more than one, select the one that has the best performance. VirtualDUB AuxSetup (comes with any version of VirtualDUB), apart from setting up frameserve, has a very useful benchmark utility that allows you to test the performance of each drive as well as identify the highest framerate and framesize you can use without dropping frames).
DEFRAGMENT THE DRIVE YOU HAVE SELECTED. OTHERWISE PERFORMANCE WILL BE LOWER

2. Use VirtualDUB capture (detailed instructions provided on the VirtualDUB site). Make sure you don't use compatibility capture but the new option using WDM drivers. (You need a recent version of VirtualDUB and if you have a P4 machine, use the P4 optimized version).

3. Capture to AVI, using DivX as a codec, setiing compression to Quality based with a value of 2 (best). Also, make sure you configure DivX to insert index frames at least twice a second (15 frames for NTSC). This gives you an average file size of 1.5 GB per hour (at least for me in PAL).

4. If you have a fast machine you can also crop and deinterlace in real-time. My machine is P4/2.8GHz and I can do that with a 30% CPU utilization during capture. If your machine is not as fast, don't deinterlace or crop during capture.

5. Audio, should be saved either in MP-3 (I use the Fraunhoffer codecs) or uncompressed. Bitrate even for uncompressed is rather low and doesn't affect throughput much.

Capture what part of the tape you want, and don't care about trimming or removing ads between content.

Trim the AVI file using VirtualDUB in direct stream mode. If you have two separate drives, read the AVI from one drive and write to the other. This way, no decompression - recompression is required and (depending on the machine and the disks) dubbing speed is fast (500 ~ 1300 fps).

Doing all that, you will end up with an AVI file to compress with any MPEG compressor. The image quality DivX produces with a Q=2 setting is perfect so it's a good source for MPEG compression. Make sure any filters you want to apply are done with VirtualDUB, they are the fastest. Also, make sure that if you apply filters, you do it once with all filters in the sequence you want. This is the faster way and causes minimum image quality degradation.

The above medhod has worked for me and I have no problems with dropped frames or sqweeks in audio. Hope it helps you as well.

[DJRumpy]

I agree. VirtualDub is excellent for Capture. I prefer XviD using 1Pass CBR @6000Kbps. I'll have to try Sasi's settings. Ya never know... 8)
Your mileage may vary. If you have enough space, and quality is paramount, consider using Huffman for your codec, instead of XviD, or DivX. It's lossless. Also, make sure you set your audio so it stays in sync if you drop any frames. A small number of dropped frames is harmless, and won't usually be noticable. If you see the dropped frames increasing by the second, your compression ratio is too high (try easier settings for your codec), or your capture resolution is too high.

Make sure you defrag the drive your going to capture to. It will have a large impact on the drives performance.

4. If you have a fast machine you can also crop and deinterlace in real-time. My machine is P4/2.8GHz and I can do that with a 30% CPU utilization during capture. If your machine is not as fast, don't deinterlace or crop during capture.

SaSi, what are you using for deinterlace? A built in, or custom filter? Are you also performing IVTC, or just deinterlace?

On the audio , go for uncompressed. It's tiny in the scale of things (usually around 1-2 GB or less). It will also reduce your CPU overhead. MAke sure you capture in the Khz of your target media (44.1 Khz for VCD/CVD/SVCD, or 48Khz for DVD). This way, you won't have to resample.

[Coop]

I probably should have clarified a bit.

I've been using Virtual VCR for captures, and it has been working great, I've completed 3 DVD projects thus far. This is the first older tape I've tried to capture, so I'm pretty sure it's the source, not the process.

For this current project, I have a total of 4 dropped frames, all at the beginning (I'm thinking that just like cassette tapes, video tape also wears harder at the beginning).

I'd like to continue my current process, which is:

capture using Virtual VCR to AVI, transcode AVI to MPEG2 using TMPGenc, author, and burn to DVD. Ideally, I'd like to strip the audio and video out before converting to MPEG2.

[SaSi]

DJRumpy,

I use VirtualDUB's internal de-interlace (in Capture menu, Video --> Filters --> Deinterlace). I have found this one to be the less intrusive de-interlace filter.

Coop,

If your dropped frames are few and in the beginning, then the distortion in the audio is probably caused by other problems. Since cleaning the heads improved things a bit, it appears that your problem may be dust-oriented.

An interesting detail about VHS audio. If your VCR audio is stereo, then the audio heads are part of the spindle with the video heads. Cleaning cleans both. If however your video is mono (or the tape was recorded on a mono VCR) then audio is recorded and read by another head that is stationary. I have also had some problems in the past by using low quality tapes that corroded the audio head. What I did was to open the VCR and clean the audio head with an cotton-tub and pure white alcohol. If your audio heads are as dirty, the tub will become brown-black.

Removing dirt is the best way to clean the audio.

If that doesn't do it, CoolEdit PRO is by far the best audio editor I have used. I have had to de-noise a soundtrack once and what I did was to use VirtualDUB to save the soundtrack to WAV, open it with Cooledit and use the noise reduction filter. Problem is CoolEdit is a beast by itself and a bitch to understand and use efficiently (I have more than 200 hours of work with Cooledit) but for someone using it for the first time, I can only refer you to my own despair when I started).

You can then re-mux the soundtrack with the video in VirtualDUB by opening the same AVI and ignore the audio part. Use WAV audio and select the cleaned WAV file.

Hope this response is closer to the problem you have.

[DJRumpy]

SaSi, I'll have to give it a try. Does it use BOB? How's the quality? I've been using AVISyth to IVTC, but I haven't been to happy with the auto detection, as they always seem to miss, and you end up with a jumped frame. I may try just deinterlacing, leaving the source at 29.97 instead. Suggested settings?

[Coop]

SaSi,

I cleaned the entire VCR when I cleaned the video head, the audio remains scratchy but it does sound better than it did. This is the first older video I've tried, and given that the 3 other captures I did from newer tapes went almost perfect, I'm sure it's just the age of the tape.

The VCR has AGC, which I think is causing the "scratchy" sound as it tries to adjust - I may just have to live with it for this project, I'm not sure how much filtering I can do without quality loss.

[SaSi]

DJRumpy,

for de-interlace during capture, I use the "best" option VirtualDUB offers, which is blending of fields. It can also discard and duplicate field 1 or 2, however these settings are not de-interlace settings but film destruction settings .

I have (some months ago) visited a site (can't remember the url but think I have saved the page on my disk - will try to locate it) that had an in-depth discussion about interlacing. It said that unless de-interlacing blends and takes horizontal motion into account, it destroys the quality. Well, risking to be blamed of praising Avery Lee, the blending he does in de-interlacing seems to take motion between fields of the frame into account. And, I have also found out that de-interlaced material compresses better.

You can de-interlace after the capture (but must make another pass). I de-interlace during capture because the CPU can take it and because I capture a whole tape overnight (I have a SP/LP/EP VCR and some of the tapes have more than 6 hours of content). So, I set VirtualDUB to stop capturing after as many hours the tape lasts, while at the same time de-interlacing, cropping sync lines and doing some noise filtering ("for free"). In the evening, I just cut crap out of the content I need and keep the streams I need (in direct stream copy @ 1000~1200 fps).

[DJRumpy]

Compressing a new clip as we speak. I'll have to see what kind of out put I get, as far as compression. The quality seems good though, will very few artifacts. My progressive scan player, HDTV, and interlaced captures don't seem to get along very well. It could be that my last cap had the field order wrong (mine's top first, but I always forget to check it). I'll have to try both methods to see if the artifacts I was seeing were due to interlaced material on a progressive display, or simple stupidity on my part.