HOWTO: Install and use the latest FFmpeg and x264

[mc4man]

Hi mc4man,
A little aside from the main discussion, is that the default kernel for natty

As of the other day, I gather it will use 2.38 eventually.
Overall I'm not too sure i'll have any use for 11.04 on my recent laptop, though I do like it for a much older p4 desktop with 23" display (where i have it now.
Probably will redo the eventual 2.38 when and if a bfs patch becomes avail., I have seen a bit of overall improvement with that on the p4 in 10.10 (2.36

[ch3rryc0ke]

EDIT: Fixed my own problem

I had inadvertently entered the svc command twice, which created another ffmpeg src folder nested in my ffmpeg source folder, causing the problem. Deleted everything and recompiled and all is well.

------

Thanks a lot for the guide..

I'm getting the following error on Ubuntu 10.10 when running the make command for ffmpeg. This seemed to happen towards the tail end of compilation as it took a good 10 minutes of compiling before I hit this error.

Also, there were no problems with the configure step.

I believe the version of ffmpeg is 26326.

Any idea what it could be? What should I do now to get it working?

AR	libavutil/libavutil.a
CC	ffmpeg.o
ffmpeg.c: In function ‘term_exit’:
ffmpeg.c:424: warning: zero-length gnu_printf format string
ffmpeg.c: In function ‘opt_audio_sample_fmt’:
ffmpeg.c:2891: warning: passing argument 1 of ‘list_fmts’ from incompatible pointer type
cmdutils.h:165: note: expected ‘void (*)(char *, int,  int)’ but argument is of type ‘char * (*)(char *, int,  enum AVSampleFormat)’
ffmpeg.c: At top level:
ffmpeg.c:247: warning: ‘sws_flags’ defined but not used
ffmpeg.c: In function ‘output_packet’:
ffmpeg.c:1045: warning: dereferencing pointer ‘picture2’ does break strict-aliasing rules
ffmpeg.c:1603: note: initialized from here
CC	cmdutils.o
cmdutils.c: In function ‘init_opts’:
cmdutils.c:67: warning: ‘sws_getContext’ is deprecated (declared at libswscale/swscale.h:191)
cmdutils.c: In function ‘read_file’:
cmdutils.c:735: warning: ignoring return value of ‘fread’, declared with attribute warn_unused_result
LD	ffmpeg_g
CP	ffmpeg
STRIP	ffmpeg
strip: Warning: 'ffmpeg' is not an ordinary file
make: *** [ffmpeg] Error 1

[FakeOutdoorsman]

Yesterday I updated the Meerkat guide to Git because that's what FFmpeg is now using, but the Git repository location changed since then. If you followed this guide yesterday and plan to use the ffmpeg directory to update FFmpeg then issue the following command:

cd ~/ffmpeg && git remote set-url origin git://git.ffmpeg.org/ffmpeg.git

[Amefurashi]

Hello,
sorry for the very n00b question: I succeded in compiling x264 on my Maverick Merkaat 64-bit using the guide in the first post, but I only get the 64-bit build. Is there a quick method to obtain the 32-bit build too?
Also, are there any differences between "make" and "make fprofiled"? Does the second option provide a more optimized code?
Thank you!

Regards,
Ame

[qyot27]

Hello,
sorry for the very n00b question: I succeded in compiling x264 on my Maverick Merkaat 64-bit using the guide in the first post, but I only get the 64-bit build. Is there a quick method to obtain the 32-bit build too?
Also, are there any differences between "make" and "make fprofiled"? Does the second option provide a more optimized code?
Thank you!

Regards,
Ame

Personally I think doing a 32-bit build on 64-bit Linux is a bit useless*, but it should be a simple matter of using --cross-prefix and --host, specifying the 32-bit architecture (what is it, i686-linux-gnu or something like that? check in /usr).

*as the relevant tools on Linux are all usable in 64-bit for a proper toolchain. The only exception would be if you were using a more powerful 64-bit-running machine to compile software for an older 32-bit-running machine. Then you'd need to use --prefix so as not to end up installing in an inconvenient spot in the file hierarchy.

If you want to build ffmpeg as 32-bit, you'd need to use --cross-prefix, --target-os, and --arch (or maybe just --arch, but it may complain about not having the other two).


I usually do cross-compiles for Windows using mingw, but it's the same basic principle to do it for any other given system, provided the right cross-compiler is installed (and on a 64-bit system, the 32-bit compiler is a cross-compiler). In my case, x264 looks like this:

PKG_CONFIG_PATH=$HOME/win32_build/lib/pkgconfig ./configure --prefix=$HOME/win32_build --cross-prefix=i586-mingw32msvc- --extra-cflags="-I$HOME/win32_build/include -march=pentium3" --extra-ldflags="-L$HOME/win32_build/lib" --qtsdk=$HOME/qtsdk --host=i586-mingw32msvc

Some of that extra stuff is because I include LAVF, FFMS2, and Audio support; look at how I specified --cross-prefix and --host. The hyphen at the end of the cross-prefix command is necessary, because it will try to look for the regular compiling toolchain with i586-mingw32msvc- attached to the front of it. So instead of looking for gcc, it looks for i586-mingw32msvc-gcc. Without the hyphen, the needed calls would be misspelled.

So in the case of 32-bit Linux, it might look like this instead:

./configure --cross-prefix=i686-linux-gnu- --host=i686-linux-gnu

[Amefurashi]

The only exception would be if you were using a more powerful 64-bit-running machine to compile software for an older 32-bit-running machine. Then you'd need to use --prefix so as not to end up installing in an inconvenient spot in the file hierarchy.

Indeed, that's exactly my purpose: providing 32-bit builds (not for personal use), but compiling them on my 64-bit maverick.

Also, I'd love to implement LAVF and FFMS2 support in my builds: is it possible under linux?

Thanks again,
Ame

[andrew.46]

Hi Ame,

x264 should automagically enable lavf input for you in the presence of FFmpeg header. This should be seen with ./configure as follows:

andrew@skamandros~/Desktop/x264$ ./configure
Platform:   X86
System:     LINUX
asm:        yes
avs:        no
lavf:       yes
ffms:       no
gpac:       no
gpl:        yes
thread:     posix
filters:    resize crop select_every 
debug:      no
gprof:      no
PIC:        no
shared:     no
visualize:  no
bit depth:  8

You can run 'make' or 'make fprofiled' now.

and after compilation should show:

andrew@skamandros~/Desktop/x264$ x264 --help | head -n 14
x264 core:112 r156 22bfd31
Syntax: x264 [options] -o outfile infile

Infile can be raw (in which case resolution is required),
  or YUV4MPEG (*.y4m),
  or Avisynth if compiled with support (no).
  or libav* formats if compiled with lavf support (yes) or ffms support (no).
Outfile type is selected by filename:
 .264 -> Raw bytestream
 .mkv -> Matroska
 .flv -> Flash Video
 .mp4 -> MP4 if compiled with GPAC support (no)
Output bit depth: 8 (configured at compile time)

Andrew

[LyingB]

I think you should remove subversion from Updating FFmpeg and x264 section.

[FakeOutdoorsman]

You're right. I missed it when I git-ized the guide. Thanks.

[qyot27]

Also, I'd love to implement LAVF and FFMS2 support in my builds: is it possible under linux?

To trim it down to size, the basic outline looks like this:

FFmpeg (stripped down), installed to custom prefix
FFMS2, installed to custom prefix
x264, LAVF and FFMS2 should be detected and used
(if you want a fully-featured FFmpeg you'll then need to build it again as this thread's main guide details)

Just copy-and-pasting from my most recent reference file, with some adjustments:

git clone git://git.ffmpeg.org/ffmpeg.git
svn checkout http://ffmpegsource.googlecode.com/svn/trunk/ ffms2
git clone git://git.videolan.org/x264.git

FFmpeg (basic):

cd ffmpeg
./configure --prefix=$HOME/ffms2_build --enable-gpl --enable-version3 --enable-postproc --disable-encoders \
--disable-muxers --disable-debug --disable-network --disable-hwaccels --disable-indevs --disable-outdevs \
--extra-cflags="-march=native"
make
make install

FFMS2:

cd ../ffms2
./configure --prefix=$HOME/ffms2_build PKG_CONFIG_PATH=$HOME/ffms2_build/lib/pkgconfig
make
make install

x264:

cd ../x264
PKG_CONFIG_PATH=$HOME/ffms2_build/lib/pkgconfig ./configure --prefix=$HOME/x264_build --extra-cflags="-march=native"
make
make install
make distclean
PKG_CONFIG_PATH=$HOME/ffms2_build/lib/pkgconfig ./configure --prefix=$HOME/x264_build/x264-10bit \
--extra-cflags="-march=native" --bit-depth=10
make
make install

Of course, adding in the proper options for cross-compiling (the --extra-cflags="-march=native" parameter should be changed to whatever processor the target machine uses*; if that's not known, leave the option out entirely). The 10-bit step for x264, starting with the 'make distclean', can be omitted if you don't find it necessary.

*actually using 'native' will set it to the processor of the computer doing the compiling. The reference file actually has pentium3 specified since I'm using a Coppermine-based Celeron.