Very low HDD space: which capture codec?

[Chainmax]

I am planning to buy a capture card (a Compro Videomate Ultra) and start capturing soon. Thing is, I'll have very little HDD space to spare (10 GB) for a while. That shouldn't be that much of a problem since I'll only be capturing ~15min cartoon episodes for now, but then again the captures will be done at full resolution. I wanted to know if capturing with FFDShow's huffyuv with the following settings:

- YV12 colorspace.
- Median Predictor.
- Adaptive huffman tables.
- Assume interlaced if height > 288 pixels

would be a good idea. If it is, I'd like to know how do you feel it would compare to MJPEG at quality 18 and 19. Thanks in advance.

[JoeShrubbery]

Personally I use FFDShow's mpeg4 for capture. Setting the min/max i-frame interval to 1, disabling b-frames, dropping all the motion estimation settings to their lowest levels (and thus quickest settings), and setting compression to one-pass constant quantizer at 2 works beautifully for me. 100% keyframe for clean editing, while minimizing demands on the CPU and giving it a plenty high bitrate considering the format, if you're gonna go to lossy captures in the yv12 colourspace then I'd say you'd be hard-pressed to find a better arrangement. I get about 5 gigs per half hour of recording time with this, although it ultimately depends on how noisy your source is (noisier source=larger capture files), plus as I'm here in NTSC-land and am uncertain as to how well my ballpark figure might hold up if you happen to be dealing with PAL material.

[Raziel6969]

Hello,

I've making some testing with anime, I like:
Lagarith 1.36, good balance in speed and compression
FFV1 AC (keyframes distance=1), best compression and good 4 editing
FFV1 AC (default: 250 keyf.dist.), bestest compression

Note: all these codecs are lossless. best quality
For capture with quality loss, PICvideo MJPEG is fast.
But is no good idea to capture to a lossy codec.

I capture to MJPEG to 100% when i have little space hdd
because my capture card has MJPEG compession in hardware.

I hope it helps. Bye

[azsd]

Alparysoft Lossless Video Codec@realtime compress level
faster speed and lower system load than FFV1/MSU Lossless/HuffYUV in my option

[mg262]

About a year ago I went through all the lossless codecs I could find testing for speed and compressability... and I found that Alparysoft Lossless Video Codec wasn't lossless. I.e. when I uncompressed the output to raw and compared with the original using fc, there were differences. I do remember that I checked my settings quite carefully and repeated the test.

Now it might well be that this was a bug with one version that has now been fixed... but I would advise anyone trying it to compress/uncompress and check against the original. (+Even if there are differences they might be too small to matter for capturing purposes.. but IMO again worth checking.)

Edit: @Chainmax, FWIW I normally capture with normal HuffYUV YUY2 and IIRC reckon on between 500 and 700 MB/min. To be more concrete, I have some 25 minute <-VHS cartoon captures on my hard disk and they all are around 15 GB. So you would be cutting it tight with normal HuffYUV but given that you're using the better compressing one inside ffdshow and also YV12, you'd probably be okay. (I would recompress one of those episodes for you to be more concrete but I don't have VFW ffdshow at my system at present -- sorry!)

[mic]

Nero actually does a passible job, or what you might consider a passible job, recording direct to DVD. If you've got a DVD writer, won't cost anything to give it a try.

[Mug Funky]

plus as I'm here in NTSC-land and am uncertain as to how well my ballpark figure might hold up if you happen to be dealing with PAL material.

480*29.97 = 14385.6
576*25 = 14400

so theoretically PAL should be very slightly more demanding, if the capture width is the same (704 is a good width to cap at IMHO)

though for DV capture, the data rates are identical for NTSC and PAL (because it must be strictly CBR to be written to tape at a constant speed).