Help me choose a cheap PCIE or USB analog SD capture device for VirtualDub

[jagabo]

You may need to have the Pro version to get all I frames (I don't have it installed anymore). I know the fastest settings require the Pro version. I should have mentioned that.

[PuzZLeR]

It's up to you, but I still encourage you to go lossless if this content is precious to you.

For giggles, I would love to personally test I-frame Xvid, and would love an alternate compromise for lesser captures. I have faith it will be impressive, and will still be amazing in quality/bitrate, but regardless, it will be inferior to 4:2:2 lossless caps.

Keep in mind, even a target quantizer of 1 is still lossy. I'm not sure if Xvid has 0, but that would be a theoretical lossless value.

Also, the reason you're not seeing much difference de-interlacing is because it deteriorates video and throws out much information. This is with any Source. If you intend to make this your new Source (disposing of the interlaced caps - not recommended) then you're fine using Xvid.

But I'm still curious to test motion with I-frame Xvid nevertheless.

[PuzZLeR]

I tried Divx as well, but I couldn't find an all-I setting, do you know how to achieve that? (if it is possible with the non-PRO version)

Pro Version only confirmed. (Wow, memories of my days being a DivX Forum Mod. )

[PuzZLeR]

Ok, ran a small test. (Probably getting late in crespo80's time).

Pointing out, that I use a dedicated capture box, an older XP PC that is nearing retirement mode. Unfortunately, it's slower, and drops frames on the more CPU intensive Xvid caps (but not on HuffYUV), so couldn't provide a good set of clips.

It would take me forever to set up one of my faster PCs for this, however I don't think it's necessary. It's clear regardless that the HuffYUV captures came out a bit cleaner, with less artifacts (but HuffYUV was almost twice the size, and you can only tell under more magnified conditions).

The panning is smoother with HuffYUV (follow the sideline in the attached clip). However, I'm not sure if it's a decoding issue, or the fact that my capture dropped and inserted a few frames with Xvid, so not sure yet on that one.

I took a frame from a small clip from a (North American) football game in a fast, and panning, scene. Forgive the quality, the tape has been recorded, re-recorded, and abused, dozens of times, but hopefully it helps.

HuffYUV cap


Xvid cap (with crespo80's settings):


My conclusion, for all it's worth, based on these clips? I wouldn't mind using Xvid at all. The differences are minor. However, I would only use it for less important stuff. After all, it seems silly to me to devote huge file sizes to (for example) a grainy quality sitcom from an older cable box recorded some decade ago.

[PuzZLeR]

Have to note: to any parties that may be concerned - no copyright infringement is intended. This was only for demonstration purposes. I will remove it if asked.

[jagabo]

Did you set Xvid to interlace mode?

<edit>

Oh yes, I see crespo80 specified interlaced mode.

[jagabo]

One other thing worth mentioning if you are going to use Xvid to caputure: enable its deblocking and deringing features. That will reduce the DCT ringing artifacts a bit. The settings are in Other Options -> Decoder.

[crespo80]

@PuzZLeR
Your xvid encode stutters because frames 25 and 26 are identical (you're probably right your capture dropped and inserted frames).
It's hard to judge the compression quality of different codecs during different captures, because the noise of an analogue capture is always different, so every pixel has a slighty different color everytime.
I think it's better to compress an already captured video: doing so you will find that, by using only your naked eye, the xvid capture is identical to the huffyuv YV12 (and almost identical even to the YUY2 version) in every single pixel, even when zoomed in, and the ringing artifacts become only visible if you strongly contrast the image, as jagabo correctly pointed out.


@jagabo
correct me if I'm wrong, de-blocking and de-ringing are only post-processing options of the xvid decoder (not the encoder), so nothing changes while encoding, it just applies those effects while decoding; but not in my case becuase I use the Divx codec to decode the video.
I tried to encode in Divx TQ1 all-I mpeg-matrix interlaced (by enabling the trial Pro period) and while it' a 20% faster than xvid, the final video is almost 10% larger and I can see it's different from the source, while the xvid is identical (at the naked eye).

[jagabo]

correct me if I'm wrong, de-blocking and de-ringing are only post-processing options of the xvid decoder (not the encoder), so nothing changes while encoding, it just applies those effects while decoding;

Yes, those filters are used during decompression. I assumed you were using Xvid to decode since you were using it to encode.

but not in my case becuase I use the Divx codec to decode the video.

Divx has similar filters.

[PuzZLeR]

Your xvid encode stutters because frames 25 and 26 are identical (you're probably right your capture dropped and inserted frames).

Yes indeed. Had to rush off after my last post so didn't have time to check, but did see VirtualDub reported that there were inserted frames (quite a few of them). This is why I made it clear it likely was my gear, not the fault of Xvid.

I use a 6 year old PC dedicated to only captures (mostly SD TV and VHS caps), which has been adequate and frees my other PCs. However, if I want to capture in Xvid, I would have to redo my setup since it has trouble handling it with those settings. My apologies.

It's hard to judge the compression quality of different codecs during different captures, because the noise of an analogue capture is always different, so every pixel has a slighty different color everytime.

Happens all the time, even on my two clips here. You can tell there were different colors and grain levels, even tracking and calibration were in different modes, and it's like that no matter what I do. VHS video is not perfect at all - I believe even the weather plays a role. (Oh, how I always hated the format all these years. So glad it's pretty much behind me now.)

I think it's better to compress an already captured video: doing so you will find that, by using only your naked eye, the xvid capture is identical to the huffyuv YV12 (and almost identical even to the YUY2 version) in every single pixel, even when zoomed in, and the ringing artifacts become only visible if you strongly contrast the image, as jagabo correctly pointed out.

As another test, I took my HuffYUV cap and encoded that to Xvid using similar settings (I frame only, etc). It was nearly identical for half the size of the HuffYUV cap. Unless my eyes are no good, I see no issue with the motion and panning either.

Having said that, I do believe this is the way to go if you wish to archive to DV, high bitrate MPEG or MPEG-4 ASP (such as DivX, Xvid), or even AVC, and this would be regardless of my slower PC that has difficulty capping directly to Xvid. Unless I missed something, the quality always seems better starting with lossless to me.

Based on this test, archiving to I-frame Xvid seems like a good choice after the lossless cap.