Doom10 Forum: Digital Video Discussion

There is no page with an exhaustive statistical comparison of every option.  That sounds like far too much work and far too much data to make sense of.

Anyway, Dark Shikari has said that subme is the biggest knob to turn (after the actual presets, you really should use them) in terms of the effect it has.  I don't remember his actual words (or where he said them) but he said, to the effect that he would increase subme before increasing any other option.

The exact settings used by each preset should give you an idea of how effective a particular one is.  For instance, merange is only increased in placebo meaning that its effect on speed is great while its return in quality is small.

There is no limit on consecutive B-frames and the level you set will take care of reference frames (assuming you're using x264.exe).  Look at http://www.x264bluray.com/ to be sure that you get everything right.

Are you using the command line encoder?  If so I have no idea how you managed to make it say that.

If you're the same adi as appeared on IRC for a few minutes this morning, why did you use the word "implement"?  It sounds like you are trying to use the API to make your own encoder.  In that case, of course it wouldn't recognise x264's command line arguments.  You'd have to make it do that.

For 8-bit crf 0 and qp 0 are the same.  In general qp 0 is now preferred for enabling lossless as it is the same regardless of the output bit depth.

No and No.

What do you mean by '"meaning"'?  You will be able to recover the exact bits that were given to the encoder, this is what lossless means.  If you resize, filter, or in any other way alter the video before encoding you will not be able to do that.

Well I see a couple or resizes and several converts in there so I will need you to post the exact script you're using.

I don't know if it exists in avxsynth but, put an Info() at the very end of the script and see what it printed on the video in mplayer.

Why are you reading from stdin?  Where is the video coming from?  What format is it being sent in?

Could it be, that this "ffms support (no)" from the "x264 --help"-output is indicating the problem?

No.  x264 does not use that for y4m files.  I see that you have compiled with lavf, so why don't you try giving the mov file as x264's input file and skip the decoding step with ffmpeg?

However it may be helpful if you could post the first "line" of the y4m file that you've made.  Running "head -n1 FILE.y4m" should do the trick.  If you could post that here we might be able to spot where the problem is coming from.

You have made a y4m file but told x264 that it is raw.

No to all your questions.

Do you mean "should be crf or 2-pass"?  If so then yes, that is what is recommended.  Since crf is the closest thing we have to constant quality it does produce a varying bitrate throughout a video and across different videos which makes it VBR.  This makes the first part of your post wrong.  VBR is definitely recommended.

You need to use a tool which can calculate the PSNR by comparing the source video and the decoded video.  A small one comes with ffmpeg.  MSU has another.  Avisynth has a filter which will print the value for each frame onto it.

Yes, you can change to a new file at an IDR frame (without open GOPs).

That looks like a simple levels problem and is caused by playback not encoding.  But after looking at some histograms I'm not so certain.  So...  How did you make these two images?  Can you read the script that the one click encoder uses?  What about the encoding settings?

Avisynth is getting RGB video because something in DirectShow is converting to RGB.  Without more information I can't tell you what to do.  You need to check what filters are being loaded with something like Graph Studio and then check what filter is turning the video into RGB.  If you're using ffdshow, check that you have allowed the YUV output you want.

Because the fast first pass disables the 8x8 transform which then makes the stream Main Profile.  Why all this bother over a simple warning from x264?  You don't need the video from the first pass.  If you had just done the second pass, you would have got a High Profile stream and no warning.