Automating Capturing and Encoding to MPEG4?

[Apulo]

I've sorted through my tape archive and extracted those tapes that contain material that is still of value, and that cannot be found anywhere else in any format. Most of this is newsreports, documentaries and obscure movies that were once broadcasted on TV.

This still turns out to be a fairly large number of tapes (240min. per tape, all mostly filled). I'm left with approximately 160 tapes (out of about 800). 70% VHS, 30% SVHS. All of it in PAL format.

These recordings are all directly off cable or antenna, and fairly good quality. A bit of noise filtering may make them better to compress, but no drastic restoration measures seem necessary.

(Apart from a few really bad ones, I'll have to see what I do with those - multigeneration copies on LP SVHS, very noisy and unstable, yet irreplacable material.)

So that's about 150 tapes with various newsreels, reports, documentaries that need encoding. An enormous amount of work, if done all by hand. So I'd like to limit the manual labour to a minimum. Luckily, almost all of this stuff is commercials-free, as it stems from channels that don't pollute broadcasts with those . So all that needs to be done is to cut at the start and ending, no further editing is needed.

I'd like to hear about possibilities/tools to automate/script the preparation and encoding in MPEG4 (DIVX, Xvid, Nero?), and would much like to hear advice on which filters to use, what codec would be best etc. I'd like to get the best result possible with this automated process, fully realising that the very best picture quality would require tweaking by hand, so I'm willing to compromise here.

Also, how to capture? I have at my disposal a nice Terratec TV card with 10bit Philips capture chip, and also an ADVC100 from Canopus (but I read at Slashcam you need to make sure the encoder is compatible with Canopus because they aren't fully DV codec compliant?)

I'm starting from the premise here that I capture in high-bitrate MPEG2, a lossless codec or DV, and then software encode to MPEG4, but perhaps direct capture to DIVX with a Plextor capture box would be better? These boxes are now sold off for 80 Euro (the one without TV tuner).

Any ideas on how to tackle this large workload efficiently would be very much appreciated!

[Brian Ellis]

First of all, why are you seeking quality when you don't have quality to start with? VHS will never give even anything approaching justifying high bitrates and S-VHS is not that much better. And you can't automate changing the tapes if you have an ordinary player.

IMHO, you already have the best hardware for the job if you seek minimum hassle. Use the ADVC-100. The question is how to capture. I suggest you may care to download the MSP8 trial; this will allow you to transcode and trim into, say, 4500 kbit/s MPEG-2 (no need to go higher from a S-VHS source, as the quality won't improve) and just use that. Don't encode into MPEG-4, you will lose quality. There will shortly be a low-cost MPEG-4 plug-in for MSP8, but I don't know whether it will allow direct transcoding from the DV stream.

If you wish, at all costs, to use MPEG-4, then I suggest you capture DV AVI and then encode if the MSP8 MPEG-4 plug-in doesn't allow direct transcoding.

[Apulo]

Thanks for your comments, Brian. I know that VHS isn't much to start with, but I don't want to end up with even lower quality. That's what I mean by "quality" - a relative thing, not absolute, otherwise none of us except professionals would be working with "quality". Nor do I want to automatically swap tapes, I know the capturing process will be mostly manual work. Editing the start and finish of the pieces will also be manually done. For capturing I'm more interested in advice on the hardware and the codec used.

The noisier and lower quality the source, the higher bitrates you would need in order to avoid artefacting. Or you need a very effective filtering technique. So that's another point I'd like to hear about.

Once files have been captured and readied for encoding, I do hope to automate this process as much as possible. It would be ideal if the encoder would filter intelligently rather than use fixed settings, but I'm not sure that's possible.

I much prefer MPEG4. MPEG2 takes up too much space for the same quality. That is, if the MPEG4 encoder works properly and the right codec is used. I've seen some exceptionally good MPEG4 files, that were recorded from TV in Japan, so I do disagree that it would be worse than MPEG2. I suspect however it requires a great deal of expert knowledge at this moment to make a good MPEG4 encoding.

And what MPEG4 codec would be best? DIVX, XVID, Nero Digital? Or go to the H264 codec right away?

[Brian Ellis]

Firstly, the point that I was making was that "on the fly" transcoding almost always means poorer quality, ESPECIALLY in MPEG-2. Encoding such an MPEG-2 file into MPEG-4 will mean a further drop in quality. Much better to capture in, say, DV AVI and then encode into MPEG-4, if that's your thing.

Then, noise reduction is a double-edged sword: it also reduces the resolution of an image, so you would have to be very circumspect in its use, essentially a compromise. When I capture from VHS, I don't go for NR but I don't have larger files because I use CBR for encoding, as a rule. I agree that a noisy source will make VBR think that a subject is fast moving.

You may wish to think about the ADVC-300 if you are worried about noise or have old tapes that may need TBC.

[Apulo]

Hi Brian, thanks for taking the time to respond once more.

OK, so capture is best in DV. Then encode to MPEG4. True, noise reduction will reduce resolution a bit, that's why I'm wary on using it in an automated process. If you can do trial runs and tweak things, I'm sure it will improve quality, but doing this for each file individually would cost too much time.

So you're saying just drop NR completely?

I'll give the ADVC300 a look (sounds expensive though )

What I am a bit worried about is the Canopus codec compatibility problem that got mentioned on Slashcam (http://www.slashcam.de/artikel/Grund...terschied.html)

They state if a Canopus file is exported to MPEG the encoder should be Canopus-compatible. Canopus supposedly uses 256 values instead of 220 for chrominance, which causes problems for encoders.

Is this 'problem' present in the DV files captured with the Canopus hardware? Or only if I use the Canopus software codec to decode? Should I use another codec to decode, like MainConcept? It's not clear to me from the articel when this happens and when it is a problem.

[Brian Ellis]

DV is DV is DV. Canopus cannot corrupt a DV stream. Just remember that "capturing" a DV stream to a DV AVI is not capturing; it is a simple copying to HDD. The only thing the codec does is to add the header, which has no influence on the file content.

I've been using ADVC-100 generated conversion to making MPEG-2 for years without problem.