avs4x26x v0.10.0 ( support high-bitdepth avs & customized x264/x265 path & avs+ )

[06_taro]

Hmm, it seems that x264-audio uses lavf as default audio demuxer, and lavf can handle pipe protocol....

[Nevilne]

Maybe something among the lines of creating 2 named pipes and feeding them to x264? Don't know how well it would work on windows though. Starting cygwin for a simple encode is kinda meh.

[06_taro]

Well, there is such a tool for windows: http://www.nmm-hd.org/newbbs/viewtopic.php?f=8&t=631
What is need to do is to create two named pipes out with one tool, such as ffmpeg/libav, then let x264-audio get those two named pipes in simultaneously, like this:

libav -i input.avs -f wav -vn \\.\pipe\1001 -f yuv4mpegpipe -an \\.\pipe\1002

x264 \\.\pipe\1001 --audiofile \\.\pipe\1002 -o output.mp4

[Chyrka]

06_taro, are there any speed difference between avs2yuv and avs4x264mod? (x64 system)

[Nevilne]

Originally Posted by 06_taro

libav -i input.avs -f wav -vn \\.\pipe\1001 -f yuv4mpegpipe -an \\.\pipe\1002

x264 \\.\pipe\1001 --audiofile \\.\pipe\1002 -o output.mp4

Am I missing something or it only allows one standard input pipe?
Like soranamedpipe \\.\pipe\123 | app -i - ouput

Speaking of which, this also doesn't work, hehe
avs2pipemod -wav audio.avs | avs4x264mod video.avs --audiofile - -o output

[06_taro]

Originally Posted by Chyrka

06_taro, are there any speed difference between avs2yuv and avs4x264mod? (x64 system)

There shouldn't be any.
Here's a rough test: http://k4095-takuan.blogspot.co.uk/2...blog-post.html

Originally Posted by Nevilne

Am I missing something or it only allows one standard input pipe?
Like soranamedpipe \\.\pipe\123 | app -i - ouput

Speaking of which, this also doesn't work, hehe
avs2pipemod -wav audio.avs | avs4x264mod video.avs --audiofile - -o output

Oops, I mis-understood how soranamedpipe works.
Anyway,

avs4x264mod "video.avs" --audiofile "audio.avs" -o "output.mp4"

should work properly.

[Chyrka]

06_taro, I see. Thanks for link.

[06_taro]

Update to v0.8

FPS will be corrected to NTSC fraction if applicable. And now you can simply use

avs4x264mod "input.dgi -o "output.264"

or

avs4x264mod "input.avi" -o "output.264"

to encode, source filter will be selected and invoked according to extension of input file (requires dlls for those source filters to be loaded automatically).

[LigH]

Trying to download avs4x264mod-0.9.0-git-r62(691c5c4).7z from github, I get:

File was not found.

[06_taro]

Edited.

GitHub stopped their upload service, so please use NMM-Mirror, or wait for attachment to be approved.

[Kazuya]

Hi,
I'm trying to encode several patterns in 10bit 4:4:4 to make some comparisons between renderers, video outputs...

But, I still have an YV12 conversion warning in execution window...
Or I can convert it with ConvertToYV24 but I think YV24 is 8bit, am I wrong ?
And can't find any "ConvertToY410" working with avisynth...

Is there a way to avoid any conversion ?

Thanks !

(my image source is a handmade png 16 bit image under photoshop CS3)

My command line to encode x264 :

C:\avs4x264mod.exe --x264-binary "C:\x264\x264_10bit.exe" --fps 5 --output-csp i444 --output "E:\output.mp4" "E:\Script encodage.avs"

Execute window :

[06_taro]

(my image source is a handmade png 16 bit image under photoshop CS3)

What is the final pixel format of your AviSynth script? avs4x264mod currently only supports outputting YV12/YV16/YV24 without forcing to convert them to YV12. If the final pixel format of your script is Y8/YV411/YUY2/RGB, then ConvertToYV12() is automatically invoked.

Or I can convert it with ConvertToYV24 but I think YV24 is 8bit, am I wrong ?

Exactly. However, thanks to Cretindesalpes for his dither package, we can use up to 16-bit YUV420/422/444 formats in AviSynth. There are two formats of such high bit depth colour space: stacked and interleaved. The former contains MSB and LSB stacked vertically, so it is easy for viewing and general filtering; while the latter actually put MSB data and LSB data in the same structure that native Planar YUV formats use, so it can be used for encoding: if you tell the receiver of this format (e.g. x264) that the raw YUV is yuv420p16, then x264 will treat this fake interleaved MSB/LSB as 16-bit yuv420. You can use yuv444p10 in the similar way by:

PHP Code:

ImageSource("input.png")

# convert RGB to stacked yuv444p16
Dither_Convert_RGB_TO_YUV(output="YV24", lsb=True)

# convert stacked yuv444p16 to interleaved yuv444p16 for encoding, you can also use Dither_out()
f3kdb(Y=0, Cb=0, Cr=0, grainY=0, grainC=0, input_depth=16, input_mode=1, output_depth=16, output_mode=2) 


Then feed the script to avs4x264mod with "--input-depth=16", then such raw yuv444p16 will be passed directly to x264 without unnecessary conversion.

If your raw yuv data is neither 8-bit nor 16-bit, then vanilla x264 will firstly shift it to 16-bit then dither it back to encoding bit depth using Sierra-2-4A error diffusion dithering. If you are using x264 built with the patch Skip bit depth filter when possible, then you can even pass 10bit avs script to x264, and neither avs4x264mod nor patched x264 will do any conversion on the raw yuv:

PHP Code:

ImageSource("input.png")

# convert RGB to stacked yuv444p16
Dither_Convert_RGB_TO_YUV(output="YV24", lsb=True)

# convert stacked yuv444p16 to interleaved yuv444p10 for encoding
f3kdb(Y=0, Cb=0, Cr=0, grainY=0, grainC=0, input_depth=16, input_mode=1, output_depth=10, output_mode=2) 


Then feed the script to avs4x264mod with "--input-depth=10".

[Kazuya]

Thanks, it works very well !
And improves a lot 8bit encodes by eliminate banding.

But I think the weight of encoded file is 3 times louder than the same encode in 10bit. I only encoded patterns, so I'm note sure.

I used your first call, because my png is 16bit.
And to encode in 8bit, I only change output depth and mode :

ImageSource("input.png")
Dither_Convert_RGB_TO_YUV(output="YV24", lsb=True)
f3kdb(Y=0, Cb=0, Cr=0, grainY=0, grainC=0, input_depth=16, input_mode=1, output_depth=8, output_mode=0)

[vdcrim]

If both your input and output are 16-bit you should use a source filter that can preserve that precision, like this modified ffms. ImageSource outputs 8-bit.

FFVideoSource("input.png", cache=False, seekmode=-1, enable10bithack=True) # stacked YV24
f3kdb(Y=0, Cb=0, Cr=0, grainY=0, grainC=0, input_depth=16, input_mode=1, output_depth=16, output_mode=2)

[Kazuya]

Thanks, I will try it !

[Timothy_Gu]

I follow the link to source code and met a 404: https://github.com/astrataro/avs4x264mod

Did you delete your github account or something like that?

[06_taro]

Don't know why, works for me when I log in, but fails when I log out.
Didn't delete anything.

[nhakobian]

Github was having some file server problems yesterday, not sure if this is related or if its been fixed.

[Timothy_Gu]

Originally Posted by nhakobian

Github was having some file server problems yesterday, not sure if this is related or if its been fixed.

OK, I can see the server status at https://status.github.com/. It reads:

We're still performing maintenance on several fileserver clusters. Some repositories are currently unavailable.

I think that's the problem.

[LigH]

But it certainly has one of the most elaborate 404 pages ever. The Jedi-on-Tatooine scene even moves parallactically with the mouse position.