PDA

View Full Version : x264 2 Pass Encode CRF + Bitrate


lariva
29th September 2007, 23:04
If someone could land me a hand in this:

to get a reasonably accurate bitrate vs. quality for an encode, I do constant quality encode; the bitrate that it produces i then use with "ratetol inf" and the stats file.

The question that I have - does the crf pass need to be with the same parameters to identify "compressibility" of the frame as the second run?

i.e. does trellis affect compressibility? does the analysis / search algorithm change it? Do others?

The parameters I'm using right now (the first was the same but --crf 22)
x264_x64.exe --bitrate 3600 --ratetol inf --stats "d:\media works\temp\bhd.720p.stats" --bframes 4 -b-pyramid --ref 8 --analyse=all --direct auto --weightb --me umh --subme 6 --b-rdo --bime -8 --trellis 2 --pass 2 --threads auto --thread-input --progress --no-psnr --no-ssim --sar 27:20 --output blah

encoded 207624 frames, 4.81 fps, 3590.99 kb/s

The issue here is the 4.81 fps, I'm running on core2 e6600 @3.2 (8x2400)ghz with 1200 mhz memory set at 5-5-5-15. Cant get much faster then that nowadays. What I'm hoping to do is to eliminate the "unnecessary" parameters from the first pass to get a good stats file and reasonable estimate for bitrate but at a slightly higher speed :)

Help anyone?

Dark Shikari
29th September 2007, 23:36
No, the first pass doesn't have to be with the same parameters. Here's what I'd use for my firstpass:


x264_x64.exe --bitrate 3600 --ratetol inf --stats "d:\media works\temp\bhd.720p.stats" --bframes 4 -b-pyramid --analyse=none --direct auto --weightb --me hex --subme 5 --pass 1 --threads auto --thread-input --progress --no-psnr --no-ssim --sar 27:20 --output blah