Editing VHS captures with Sony Vegas

[KarMa]

Just putting this out there, QTGMC spatial denoises by itself automatically. Other people on this site have talked about using QTGMC just for its denoiser. So keep that in mind, you are not just temporal denoising.

I thought QTGMC's denoising effect was caused by its motion compensation, thus temporal...?

Honestly don't know, just assumed since it was interpolation it would be spatial. Now that I actually look at it, it does seem to look at adjacent frames for motion which would imply temporal. None the less it's another denoiser and should keep that in mind.

[LMotlow]

I agree. Those few artifacts are a hassle and won't go away. It looks worse if re-interlaced, no matter what you do....

A few details from various versions I tried. Below, frame 36 (original, interlaced). The usual combing, but look how blurred and soft the detail are in motion. The girl's hands on the right aren't even moving.


frame 72 (deinterlaced frame 36), QGTMC slow. Looks like motion artifacts in the boy's hands, other details that aren't moving are still blurred and indistinct.


frame 72 (deniterlaced frame 26), QGTMC slow + MDegrain2. Fingers have double edges, girl's hands are badly blurred, grain in contour details. These all look like low bitrate compression artifacts to me, yet the capture was supposedly lossless. And cloth in the chair still has too much noise, and it's not moving very much. Some artifacts on the boy's nose.


frame 72 (deinterlaced frame 36), QTGMC fast + MCTD high. Smoother overall, most of the noise in the chair is gone, girl's detail are a bit more distinct. I kept this as the best compromise.


Any idea where all the original artifacts and deinterlace problems are coming from? And yep, I tried yadif, bob, separate fields, etc. etc.

[VHS Roger]

I just wanted to update about the livelock warning I've been getting. I thought at first it was caused by something in the script, but after removing code line by line and trying different plugins I now think it is something with the x264 encoder. I don't know what settings are best to use, but after fiddling with it I now get one livelock warning at the very end of processing the video. The warning does not seem to affect the video.

[Brad]

x264vfw? Have you enabled the "VirtualDub Hack" option?

Though I don't encounter a livelock error with it disabled, merely a warning about frames possibly being lost.

EDIT:

The usual combing, but look how blurred and soft the detail are in motion. The girl's hands on the right aren't even moving.

[Attachment 32666 - Click to enlarge]

I think it's just how the camera sensor coped with low light. Maybe rudimentary noise reduction performed by the camera or EV-C100 player though.

[VHS Roger]

Yes x264vfw and I do have the "VirtualDub Hack" option enabled.

[Brad]

How about a screenshot of your "x264vfw configuration" window? (PNG is better for window screenshots than JPEG and usually smaller, BTW.)

[VHS Roger]

Here you go.

[Brad]

Perhaps set Preset to Ultrafast, tick Fast Decode, tick Zero Latency, and see what happens. Not that I would suggest to actually use that even if it removes the error. Otherwise, I dunno.

[_Al_]

Rate factor 5 is quite high, it gives you bitrate about 25Mbit/s.

I'd leave it to 15 and rather set --tune grain into that "Extra command line" section, that might give you bitrate about 15 Mbit/s. --tune grain tends to not take away grain.

[VHS Roger]

I tried messing with those settings and had problems with it only encoding the first half or so of the video.

Edit: Missed your reply Al. I've tried lower bitrate and it still gave me the warning. Thanks for the --tune grain tip, I'll try that.

Another question: Is there a way to fade in/fade out clips with virtualdub/avisynth without needing to render the entire clip again? I tried fadein() in avisynth and used "direct stream copy" but the resulting clip was much larger than the original and had lots of static.

[_Al_]

Avisynth serves video to VirtualDub as uncompressed, so Direct Stream Copy in VirtualDub exports that.

[LMotlow]

Avisynth decodes video to run filters, so using "direct stream copy" outputs uncompressed video. Select "full processing mode" so that you can get access to "color depth..." (set the colorspace you want for output) and "compression..." (set the compressor you want to use). Then select "fast recompress" and "Save Avi".

Render? That's not a term I've seen associated with Avisynth. Are you working with losslessly compressed media, or is it encoded video? Lossless media can be modified and recompressed. If you input lossy encoded video into Avisynth and filter the video (fades are filters), the output has to be lossslessly recompressed or lossy re-encoded.

http://www.animemusicvideos.org/guides/avtech/video3.htm

[Brad]

Select "full processing mode" so that you can get access to "color depth..." (set the colorspace you want for output) and "compression..." (set the compressor you want to use). Then select "fast recompress" and "Save Avi".

Simply choosing Fast Recompress is sufficient to ensure that there is no color conversion by VirtualDub. When you pick Normal Recompress, "Filters" is greyed out. When you pick Fast Recompress, "Color Depth" is additionally greyed out. This still doesn't prevent codecs from performing color conversion internally (luckily), as with Huffyuv's "Convert To YUY2" mode for RGB sources or your typical lossy compressor accepting YUY2 and creating YV12.