Dirac 1.0.2 released

[Kurtnoise]

Originally Posted by littleD

so what are those DSH filters for, then?

to play back raw streams (drc files) or AVI containing raw streams...

Originally Posted by Cruncher

are those both CBR encodes ?

yes, both are CBR...Here is the command lines I used:

FFmpeg -i input.avs -vcodec libdirac -b 1152000 output.drc

FFmpeg -i input.avs -vcodec libx264 -b 1152000 output.drc

Originally Posted by Cruncher

everyone should be aware Dirac has no ABR/VBR yet it's only plain CBR

depends on the tool you use... There is a quality mode, so I assume that's VBR. About ABR: if you use FFmpeg, you can tweak tunings and use all video switches you want from this tool. So, a 2passes encoding is possible in this case.

[Dark Shikari]

Originally Posted by Kurtnoise13

yes, both are CBR...Here is the command lines I used:

FFmpeg -i input.avs -vcodec libdirac -b 1152000 output.drc

FFmpeg -i input.avs -vcodec libx264 -b 1152000 output.drc

Doesn't look like CBR to me.

[Kurtnoise]

ABR, then ?

something like that :

FFmpeg -i input.avs -vcodec libdirac -b 1152000 -minrate 1152000 -maxrate 1152000 -bufsize 1000k output.drc

does it look like CBR ?

[guada2]

Kurtnoise13,

It is possible a GUI..
Thanks.

[Kurtnoise]

not from me.

[bratao]

Well, into my tests. The dirac at
>ffmpeg -i myfile.mp4 -b 1500000 -vcodec libdirac -strict -2 dirac_output.drc
Won by far the
>ffmpeg -i myfile.mp4 -b 1500000 -minrate 1500000 -maxrate 1500000 -bufsize 183k output.mp4

[Dark Shikari]

Originally Posted by bratao

Well, into my tests. The dirac at
>ffmpeg -i myfile.mp4 -b 1500000 -vcodec libdirac -strict -2 dirac_output.drc
Won by far the
>ffmpeg -i myfile.mp4 -b 1500000 -minrate 1500000 -maxrate 1500000 -bufsize 183k output.mp4

Well obviously, how do you expect x264 to do anything at such absurd restrictions? 183k VBV bufsize?! Are you insane?

Its not fair to have a test where you use totally different VBV restrictions for each of the encoders--you're not passing any at all to Dirac, nor are you verifying that Dirac's stream abides by the same restrictions you've passed to x264.

Surprise surprise, when you massively cripple x264 but don't cripple Dirac at all, Dirac looks better! Last I heard a Ford Focus can out-race a Ferrari if you lock the latter to first gear, too.

[nm]

Apparently he was comparing to ffmpeg4 (MPEG-4 ASP) instead of x264. Not that your remark wouldn't apply to that case too...

[Dark Shikari]

Originally Posted by nm

Apparently he was comparing to ffmpeg4 (MPEG-4 ASP) instead of x264. Not that your remark wouldn't apply to that case too...

Oh wow, you're right, he's not even using x264. I suspect this is even more unfair then, as libavcodec's VBV code is considerably inferior to x264's.

[Kurtnoise]

1.0.2 has been released...

This a a minor release complying with the latest Dirac
Bytestream Specification 2.2.3.

-Improved compression performance for small picture sizes.
-Improved compression performance at low bit-rates.
-Included macros to check the API version.
-Improved motion estimation.
-New 422 <-> 420 conversion utilities for interlaced video.
-New functionality added to encoder API. The programmer's
guide has been modified accordingly.
-Fixed bugs in handling DC subband.
-DirectShow filter released to be able to play back Dirac
raw bytestreams or Dirac muxed in AVI in Windows Media
Player and MPlayer Classic.
-Released the source and win32 binary of ffmpeg2dirac, a
utility that can transcode any multimedia format that can
be decoded by FFmpeg into Dirac video and Vorbis audio (if
audio exists) muxed in Ogg.
-Note that new patches for MPlayer and FFmpeg will no
longer be released because Dirac support via Dirac and
Schroedinger libraries is available in FFmpeg.

https://sourceforge.net/project/show...roup_id=102564

[`Orum]

While I still think it may be a little early to do a real comparison between x264 and a Dirac encoder, I think that when one is done, they should use "real" encoding settings. If the defaults are used in the test, and never in real world encoding, what does that prove other than you shouldn't use the default settings? The harder part will be determining exactly what is fair settings balance between such wildly different codecs...

[Dark Shikari]

Originally Posted by `Orum

While I still think it may be a little early to do a real comparison between x264 and a Dirac encoder, I think that when one is done, they should use "real" encoding settings. If the defaults are used in the test, and never in real world encoding, what does that prove other than you shouldn't use the default settings? The harder part will be determining exactly what is fair settings balance between such wildly different codecs...

IMO the fairest thing to do is just max encoding settings quality-wise. Dirac is not meant to be fast (that's Schroedinger's job), and even then neither is nearly as speed-optimized as x264, so it would be unfair to try to compare them with same-speed settings.

[lithoc]

I'm encoding some video with ffmpeg2dirac 0.1.0 on quad core

Average CPU usage is 30% . I think the encoder is not multi-threaded / efficient enough to use all CPU.
Secondly, it doesn't support .avs file

Any idea when is the features will implemented?

[nakTT]

Last time I heard, the research version is better in term of quality. Any truth in this?