Alternative to ffdshow's huffyuv yv12 for intermediate files?

[Chainmax]

I'm going to create a DVD, and since my machine is quite slow (Sempron 2200+ w/256MB DDR400 RAM) I always encode the avisynth script to an intermediate file before feeding it to CCE. Up to now I was using ffdshow's huffYUV YV12 with median predictor and adaptive huffman tables. What codecs would be comparable to it regarding speed and (most importantly) compression?

[Elic]

You can use ffdshow's ffv1 (selectable YUY2 or YV12, selectable huffman or arithmetic compression). Also try ArithYUV and Lagarith - both have arithmetic compression and fast enough in compressor mode, but slightly slower decompressors compared to ffv1.

[Chainmax]

What should compress better, huffman or arithmetic?
Also, are any of these faster than what I've been using? I seem to remember past threads in which it was said that FFV1 was really slow for encoding. I have a Sempron 2200+, so I need something somewhat efficient speedwise.

[Elic]

Chainmax
IMHO the fastest compressor is that matches colorspace of source :) because colorspace conversions eats more CPU power than difference between Huffman and arithmetic. HuffYUV works at YUY2 and thus consumes more CPU if source is in YV12; in similar case, you probably need FFV1 with settings "colorspace = YV12" and "coder type = VLC" (my favorite defaults).

Arithmetic coding is slightly better (approx half of bit per pixel?) and significantly slower than Huffman algorithm. Maybe those people who said that FFV1 is too slow, used arithmetic coding as well.

IMHO for your needs (DVD creating for Avisynth), most suitable will be codec that have maximum decompression speed, especially if you use 2-pass encoding. In my case (heavy filtering of dirty animations), exact FFV1 (YV12, VLC) is most preference and at 20% faster than any other combination of arith and/or YUY2.

[Chainmax]

ffdshow's huffyuv can be set to YV12, but I'll give FFV1 a try with the setting you use and huffman encoding. Thanks for the advice :).

[billou2000]

Is there a way to encode interlaced YV12 and having it played properly without tweakings?
Encoding is not a problem as physically YV12 progressive and YV12 interlaced is the same thing. But on decoding, colour components have to be applied on different lines. I guess a simple tag would be enough to differentiate them.
But at the moment I need to work from an avs file and convert to a 422 colour space ("converttoYUY2(interlaced=true)" to have the file decoded properly.

[WorBry]

There is a related threadover in the General Discussion section:

http://forum.doom9.org/showthread.php?p=794717#post794717

I'm thinking that the questions I posed there might be better addressed here. Specifically, what are the recommended settings for FFDShow HuffYuv for YV12 encoding. I'm assuming the 'Adapted Huffman tables' are advantageous, but I am not sure what Predictor (? Median) and FourCC to use.

Please dont strike me for a double post - I have only just noticed this thread.

Thanks.