Modern consensus on best sub-$250 lossless AVI-capable analog capture card?

[Mini-Me]

Remember legal YUV video is 16-235, decoded to RGB 0-255. Anything digitised to levels above 235 and below 16 will looked clipped when watched in a normal PC video player, but can be recovered and brought in range.

I forgot about that...after looking at a tape play on a crappy VCR just a couple minutes ago, I think there should be enough headroom to make a proc amp unnecessary.

Yes, but be prepared for the tweaking to take longer than you expect, and for the "best" results to need different settings for different scenes.

I don't bother - I either leave it as it is, or blast the lot with mdegrain3 or something based on it (sometimes adding a little noise back in - don't like things looking too clean).

But then I do everything to DV-AVI. So I can keep all the source files. And if I find a result that I'm not happy with when watching it days or weeks or months later, I can go back to the source and fix it.

This reminds me of another question. It's probably better-suited to its own thread, but I'll ask now that we've kind of touched the subject.

Huffyuv archives are out for the time being for space reasons, so my "fallback archives" need to be encoded in some lossy codec. I was originally thinking of ~10mb/s mpeg2 for these files, but I need some more opinions on whether that's the best codec choice. I can't keep the source/Huffyuv files, but it's likely the videos may be edited to make new ones in the future, e.g. to cut/paste segments into a "best of" montage.

On one hand, 4:1:1 to 4:2:0 issues aside, DV is apparently considered a great editing format. On the other hand, storing 150-180 hours of DV would still require at least two huge hard drives, not counting backups...and that's also getting in the "prohibitively expensive" area. I'd like to avoid that kind of storage requirement for now, if I can; for my archive copies, I'm looking for a bitrate more in the ballpark of 10mb/s.

I've read that mpeg2 and especially H.264 (using x264) are poor formats for editing. I've further heard that H.264 is poor with interlaced sources, but I've also heard that it preserves much more detail than mpeg2 (especially at low bitrates, but maybe even at higher bitrates). This comparison bears that out, although I'm sure it was crafted to show x264's strengths: http://mirror05.x264.nl/Dark/website/compare.html

Just between mpeg2 and H.264 for the moment, which would you suggest? Is there a significant difference between the two in terms of interlaced sources, editing difficulty, or editing lossiness (e.g. do simple cut/paste operations require full reencodes in one or the other)? Are one or both of them entirely unsuited to archiving, when someone might end up editing, e.g. to make montages? Would the answer change if we decoded the mpeg2/H.264 archive before performing edits, then reencode the final product?

In short, assuming DV just takes up too much space, would you suggest mpeg2, H.264, or some entirely different codec for archiving with the possibility of future edits? Finally (especially if mpeg2 is the answer), what's the best encoder to use? (...and the best free encoder? )

(Hrm...I really should have made this a separate post. )

Be aware that SD interlaced video typically looks far worse on a PC than a TV. Unless you're targeting PC playback, don't spend time fixing faults that no one will notice (potentially introducing artefacts along the way that people might notice!). I denoise and sharpen far more aggressively for YouTube or Facebook than for DVD. IMO you also need brighter (and more carefully controlled) levels for PC/web than for DVD.

Cheers,
David.

This is a good point. There's very little obvious noise when watching these tapes, so the primary reason for denoising would be just to make sure that whatever noise is there doesn't cause encoding artifacts. I guess I'll have to experiment to see whether the filter is even worth it.

It's not a matter of bitrate really. It's mostly a matter of what your DV decoder does with the chroma channels.

NTSC DV encodes colors with 1/4 the horizontal resolution. A 720x480 frame has a 720x480 luma channel but only 180x480 chroma channels. 180 pixels is sufficient for VHS chroma which is only around 40 lines of analog resolution across the entire width of the frame. But some DV decoders simply duplicate the chroma samples when upscaling to 360 YUV 4:2:2 or 720 RGB 4:4:4 pixels. This can lead to obvious posterization of colors (banding). Other DV decoders will interpolate the colors for smoother results. There are some examples of this somewhere around here. Found one:

https://forum.videohelp.com/threads/289684-DV-capture-quality-sucks?p=1758114&viewfull=1#post1758114

With VHS caps the noise in the luma channel will mask this problem to some extent.

There is also an issue with some hardware DV encoders that give an off-color cast, too little/much saturation, or bad levels, with their caps. And there's often no proc amp controls so you're stuck with the default settings.

I was originally thinking low encoding bitrates (from DV->mpeg2) could exacerbate the banding and result in more macroblocking...but it was a pretty uninformed guess, since I'm not really familiar with how mpeg2 encoding works. As you mentioned though, VHS chroma is so low-res (well below DV chroma res) that DV compression wouldn't introduce much banding anyway, assuming it's decoded smoothly. The banding would be more apparent given higher resolution sources though...so to ask a hypothetical question, would reencoding to low bitrate 4:2:0 mpeg2 in high-res cases exacerbate banding from 4:1:1 DV? (Or am I still confused? )

[2Bdecided]

Not quite the same, but this might be worth reading...

http://forum.doom9.org/showthread.php?t=148187&highlight=archiving

...and the batch script at the end could be very useful.


Another warning: transcoding to MPEG-2 or x264 can be very slow. You can edit either of them just fine. Whether you take a quality hit depends on how picky you are. It ought to be possible to go...
lossless > x264 > edit > x264
... at, say, ~8Mbps, and still end up with better quality than...
lossless > MPEG-2 DVD.

MPEG-2 maxed at 15Mbps should be without problems too, but that's not saving much bitrate.

I'm not comfortable with x264 interlaced yet (though maybe it's good enough now - I don't keep up!), so I'd do it progressive. That might cause a little confusion when you want it interlaced again - if you give an NLE 60p, and ask for 60i output, what you often get is 30p-in-60i by dropping every other frame, which is horrible. You want it to re-interlace properly, but sometimes it won't. It's trivial in AVIsynth, but you may not want to use that.

Try it and see. You might find the most basic option (DVD MPEG-2) already looks more than good enough.

Cheers,
David.