Updated x264vfw?

[Blue_MiSfit]

I'm ducking under my desk while I write this, but I'm wondering if there are updated x264vfw builds out there somewhere?

I'm working on tape (HDCAM-SR) ingest on a Digital Rapids station.

I can either do MPEG-2, uncompressed AVI, or use a VFW codec. I tried HuffYUV and it's working fine there. Of course, mathematically lossless isn't necessary, and I thought of x264 right away!

I did some tests on my desktop, and getting 30+fps out of x264 for 1080p material isn't difficult at all. Using equivalent to --preset veryfast I got great results :)

A quip with the x264vfw GUI is that it doesn't let you select bitrates above 20000. I can just use CRF so that's no problem, but I'd like predictability.

I also saw Digital Rapids supposedly works with DirectShow codecs as well?

Thoughts?

~MiSfit

[Keiyakusha]

ffdshow have x264 which makes possible encoding through vfw. Of course settings dialog is not fully up to date (i wasn't able to get 100% the same settings as --preset ultrafast) but still its somehow usable.
ffdshow r3342 reports the following:
http://imgur.com/sLDZv.png
Thats all I know about x264 vfw...
EDIT: maybe its not really new but there is some "updated x264" in changelog from time to time.

[MatLz]

Latest I know is here:
http://komisar.gin.by/

[Blue_MiSfit]

Cool, thanks for the info, guys!

@keiyakusha:
I know ffdshow has x264 encoding capabilities, but who knows which GIT version this uses?

I'll try both :)

~MiSfit

[Blue_MiSfit]

It looks like on my Q6600, I can get ~28fps for 1080p content at CRF14 by using --preset superfast and --no-cabac. If I turn off b-frames, I can get above 30fps, even in interlaced mode with an associated bitrate cost of ~50% (!).

The Digital Rapids box is a 3 GHz octo-core, so I will have a lot more headroom there. Obviously I want lots of headroom. Dropped frames BAD!

~MiSfit

[MatLz]

For the revision in ffdshow, it should be between r1196 and r1327.....maybe...:)

Edit: in the details version it appears a '75'...if it's the core number, my estimation mustn't be that bad: (1197+1326)/2 == 1261.5, and in my archives, r1259 has that core number:D