mestrelion / wavegain Goto Github PK
View Code? Open in Web Editor NEWReplaygain functionality for WAV audio files
License: GNU Lesser General Public License v2.1
Replaygain functionality for WAV audio files
License: GNU Lesser General Public License v2.1
ReplayGain for wave files, v1.3.2 WaveGain is an application of the ReplayGain algorithms to standard PCM wave files. Calculated gain adjustments are applied directly to the audio data, instead of just writing metadata as traditionally done for other formats like MP3, FLAC and Ogg Vorbis. The replaygain values can also be added as metadata in a custom RIFF chunk named 'gain'. This could theoretically allow WAV files to have same lossless functionality as other formats where audio data is not altered. But since no current players are aware of this "standard", the metadata is used only by WaveGain for the "--undo-gain" feature, which is lossy. About replaygain ------------------------------------------------------------------------------- Replaygain is a standard that allows compatible players to play audio files with an adjusted volume gain so they have a consistent perceived loudness. It is similar in concept to Normalization, but instead of using just the peak value to calculate the gain, it relies on RMS (sound pressure "power") values and psychoacoustics models to calculate a gain based on the perceived loudness of an audio file by a human listener. Loudness can be averaged either for each individual file, also called "Track" or "Radio" mode, where every song is adjusted to the same standard loudness; or by group, also called "Album" mode, where individual "songs" may sound louder or quieter than others, but each "album" as a whole has the same average loudness as other albums. For more information, http://en.wikipedia.org/wiki/ReplayGain http://wiki.hydrogenaudio.org/index.php?title=Replaygain http://wiki.hydrogenaudio.org/index.php?title=ReplayGain_specification The saga of a Debian package for wavegain: ------------------------------------------------------------------------------- - Current source code (in .zip format) and windows binaries are published at http://www.rarewares.org/others.php http://www.rarewares.org/files/others/ - An i386 binary .deb for version 1.2.6 (from 2005) was published at http://www.rarewares.org/debian/packages/unstable/index.php - Later, Linux Mint hosted the source, amd64 and i386 packages for v1.2.8 (2010) http://packages.linuxmint.com/pool/import/w/wavegain/ This package is a combination of current upstream source, a proper Debian packaging improved from the Mint release, and some patches to fix some (serious) Linux issues. Install how-to ------------------------------------------------------------------------------- There are 2 approaches: you can build and install a .deb package, or simply compile and install from source. For both I assume you are already at the source directory (which is not the root directory of the cloned repository) Debian method: This is OUTDATED, UNMAINTANED and NON-WORKING! Use Traditional Method instead! $ sudo apt-get install devscripts # if you don't already have debuild $ debuild -- binary && sudo dpkg --install ../wavegain_*.deb Traditional method: $ make && sudo make install # by default installs to /usr/local/bin/wavegain or $ make && sudo make install prefix=/usr # to install to /usr/bin/wavegain Currently, 64-bit binaries build fine but can generate corrupted audio on output files, so to prevent this by default it builts a 32-bit executable instead, which runs and works fine on 64-bit architectures. However, multi-arch libraries and headers might be needed as a pre-requiste for compiling. On Debian/Ubuntu this can be installed by: sudo apt install gcc-multilib Pre-compiled binaries and old versions ------------------------------------------------------------------------------- You can find pre-compiled .DEB binariy packages for i386 and amd64, as well as as some other goodies, like a mega-archive of all previous versions and a (lame and outdated) html manual in the Downloads section of GitHub: https://github.com/MestreLion/wavegain/downloads License and copyright ------------------------------------------------------------------------------- WaveGain is Copyright (c) 2002-2010 John Edwards <[email protected]>, a.k.a. John33, and several other co-authors and contributors are mentioned as Copyright owners in the source files. Originally released as LGPL-2.1 or later, although some source code simply says GPL, and some others some says GPL2+. Despite this, the project as a whole should be regarded as LGPL-2.1+, as per COPYING file. To avoid further confusion in the future, I suggest source files be standartized with a common header and re-licensed, preferably as GPL-3 or later (not LGPL). I claim no copyright over the C source code, including patches I made. The Debianization and aditional material, like this README and Makefile, is Copyright (c) 2012 Rodrigo Silva (MestreLion) Documentation (from wavegain --help) ------------------------------------------------------------------------------- Copyright (c) 2002-2010 John Edwards <[email protected]> Additional code by Magnus Holmgren, Gian-Carlo Pascutto, and Tycho Usage: wavegain [options] input.wav [...] OPTIONS -h, --help Prints this help information. -a, --album Use ReplayGain Audiophile/Album gain setting, or -r, --radio Use ReplayGain Radio/Single Track gain setting(DEFAULT). -q, --adc Apply Album based DC Offset correction. DEFAULT is Track based DC Offset correction. -p, --no_offset Do NOT apply DC Offset correction. -c, --calculate Calculates and prints gain settings, and DC Offsets BUT DOES NOT APPLY THEM - This is the DEFAULT. -x, --scale Writes scale values to stdout in the format: n.nnnnnn In Album mode it only writes the Album Scale value, and in Title mode it only writes the Title Scale values. ONLY works in Calculation mode. -y, --apply Calculates and APPLIES gain settings, and applies DC Offset correction. -w, --write Writes a 'gain' chunk into the Wave Header. Stores the scalefactor applied to the wave data as a double floating point number. Only written when gain is applied. Presence will result in file being skipped if reprocessed. (Unless '--force' or '--undo-gain' are specified.) --force Forces the reprocessing of a file that contains a 'gain' chunk and will result in the new scalefactor overwriting the existing value. --undo-gain Reads the scalefactor in the 'gain' chunk and uses the value to reverse the previously applied gain. This will NOT recreate a bit identical version of the original file, but it will be rescaled to the original level. -z, --recursive Search for files recursively, each folder as an album -l, --log Write log file.(Default filename = WGLog.txt) -f, --logfile Specify log filename. (Assumes -l if present.) -n, --noclip NO Clipping Prevention. -d, --dither X Dither output, where X = 0 for dither OFF (default). 1 for dither without Noise Shaping. 2 for dither with Light Noise Shaping. 3 for dither with Medium Noise Shaping. 4 for dither with Heavy Noise Shaping. -t, --limiter Apply 6dB Hard Limiter to output. -g, --gain X Apply additional Manual Gain adjustment in decibels, where X = any floating point number between -20.0 and +12.0. Clipping Prevention WILL be applied UNLESS '-n' is used. -s, --fast Calculates and prints gain settings - DOES NOT APPLY THEM. NOTE: This method does NOT process all samples, it only processes 200 x 16k chunks of samples. Results will NOT be as accurate as a full analysis but, with most material, will be within +/- 0.5db. Files of 8,192,000 real samples, or less, will be analysed in full. DC Offset is neither calculated nor corrected in FAST mode. -o, --stdout Write output file to stdout. FORMAT OPTIONS (One option ONLY may be used) -b, --bits X Set output sample format, where X = 1 for 8 bit unsigned PCM data. 2 for 16 bit signed PCM data. 3 for 24 bit signed PCM data. 4 for 32 bit signed PCM data. 5 for 32 bit floats. 6 for 16 bit 'aiff' format. NOTE: By default, the output file will be of the same bitwidth and type as the input file. INPUT FILES WaveGain input files may be 8, 16, 24 or 32 bit integer, or floating point wave files with 1 or 2 channels and a sample rate of 96000Hz, 88200Hz, 64000Hz, 48000Hz, 44100Hz, 32000Hz, 24000Hz, 22050Hz, 16000Hz, 12000Hz, 11025Hz or 8000Hz. 16 bit integer 'aiff' files are also supported. Wildcards (?, *) can be used in the filename, or '-' for stdin. wavegain-1.3.2
make on OSX fails
Here's the error output from a fresh clone:
https://gist.github.com/be302e2e7f303cf64a20399b7aa2dbbd
Minor issue but the address for FSF address in the COPYING file is deprecated, changed some time back.
The current address is
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
I have an audio file of the following specs:
Input File : 'outsox.wav'
Channels : 2
Sample Rate : 44100
Precision : 24-bit
Duration : 00:01:00.00 = 2646000 samples = 4500 CDDA sectors
File Size : 15.9M
Bit Rate : 2.12M
Sample Encoding: 24-bit Signed Integer PCM
When I try to normalize it with wavegain -y I get a file with the audio being chopped up. It also seems to have a pitch an octave higher than the original, which would point to metadata being wrongly interpreted...
The original file: http://bernhardwagner.net/outsox.wav
The file generated by "wavegain -y": http://bernhardwagner.net/outsox_wavegain.wav
This is wavegain v1.3.1 on Mac OSX 10.8.4, compiled from source.
Thank you
Bernhard
Tested in Ubuntu 16.04, 64bit
I use the following command with the file:
./wavegain wavegain_floting_point_exception.wav
and get:
Floating point exception
I use gdb to analysis the bug and get the below information:
gdb-peda$ set args wavegain_floting_point_exception.wav
gdb-peda$ r
Starting program: wavegain wavegain_floting_point_exception.wav
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
Program received signal SIGFPE, Arithmetic exception.
[----------------------------------registers-----------------------------------]
RAX: 0x5e00e84b
RBX: 0x7fffffffd4e0 --> 0x7fffffffd700 --> 0x7fffffffd9b0 --> 0x0
RCX: 0x60c00000bf00 --> 0x0
RDX: 0x0
RSI: 0x3
RDI: 0x0
RBP: 0x7fffffffd500 --> 0x7fffffffd540 --> 0x7fffffffd630 --> 0x7fffffffd720 --> 0x7fffffffd9d0 --> 0x4197d0 (<__libc_csu_init>: push r15)
RSP: 0x7fffffffd340 --> 0xc00000002 --> 0x0
RIP: 0x40a5cc (<wav_open+3809>: idiv rdi)
R8 : 0x61600000fd60 --> 0x0
R9 : 0x7fffffffd2d0 --> 0x5e00e84b61746164
R10: 0x7ffff7fc5780 (0x00007ffff7fc5780)
R11: 0x7ffff692bf90 --> 0xfffda370fffda09f
R12: 0xffffffffa70 --> 0x0
R13: 0x7fffffffd380 --> 0x41b58ab3
R14: 0x7fffffffd380 --> 0x41b58ab3
R15: 0x0
EFLAGS: 0x10206 (carry PARITY adjust zero sign trap INTERRUPT direction overflow)
[-------------------------------------code-------------------------------------]
0x40a5c0 <wav_open+3797>: imul edx,DWORD PTR [rbp-0x19c]
0x40a5c7 <wav_open+3804>: movsxd rdi,edx
0x40a5ca <wav_open+3807>: cqo
=> 0x40a5cc <wav_open+3809>: idiv rdi
0x40a5cf <wav_open+3812>: mov rcx,rax
0x40a5d2 <wav_open+3815>: mov rax,QWORD PTR [rbp-0x1b0]
0x40a5d9 <wav_open+3822>: add rax,0x10
0x40a5dd <wav_open+3826>: mov rdx,rax
[------------------------------------stack-------------------------------------]
0000| 0x7fffffffd340 --> 0xc00000002 --> 0x0
0008| 0x7fffffffd348 --> 0x60200000eff0 --> 0xff042546464952
0016| 0x7fffffffd350 --> 0x60c00000bf80 --> 0x40a8be (<wav_read>: push rbp)
0024| 0x7fffffffd358 --> 0x61600000fc80 --> 0xbebebebefbad2488
0032| 0x7fffffffd360 --> 0x100000000 --> 0x0
0040| 0x7fffffffd368 --> 0x60400000dfd0 --> 0xbebebebe00080000
0048| 0x7fffffffd370 --> 0x24 ('$')
0056| 0x7fffffffd378 --> 0x0
[------------------------------------------------------------------------------]
Legend: code, data, rodata, value
Stopped reason: SIGFPE
0x000000000040a5cc in wav_open (in=0x61600000fc80, opt=0x60c00000bf80, oldbuf=0x60200000eff0 "RIFF%\004\377", buflen=0xc) at audio.c:790
790 opt->total_samples_per_channel = len/(format.channels*samplesize);
gdb-peda$ bt
#0 0x000000000040a5cc in wav_open (in=0x61600000fc80, opt=0x60c00000bf80, oldbuf=0x60200000eff0 "RIFF%\004\377", buflen=0xc) at audio.c:790
#1 0x0000000000407833 in open_audio_file (in=0x61600000fc80, opt=0x60c00000bf80) at audio.c:362
#2 0x0000000000414ded in get_gain (filename=0x60700000df40 "wavegain_floting_point_exception.wav", track_peak=0x60600000efd8, track_gain=0x60600000efd0, dc_offset=0x60600000efe0,
offset=0x60600000eff0, settings=0x7fffffffd8f0) at wavegain.c:181
#3 0x000000000041198d in process_files (file_list=0x60600000efc0, settings=0x7fffffffd8f0, dir=0x41f1a0 ".") at main.c:197
#4 0x000000000041453f in main (argc=0x2, argv=0x7fffffffdab8) at main.c:729
#5 0x00007ffff67b7830 in __libc_start_main (main=0x4139ca <main>, argc=0x2, argv=0x7fffffffdab8, init=<optimized out>, fini=<optimized out>, rtld_fini=<optimized out>, stack_end=0x7fffffffdaa8)
at ../csu/libc-start.c:291
#6 0x0000000000401ce9 in _start ()
@MestreLion Since 1.3.1, you've added a number of really critical improvements; in particular, the issue with temporary file names tripped me up when using that version, and is now corrected in your master. Since your code has been stable for a while, could you increment the version number to encourage distro maintainers to use it?
Tested in Ubuntu 16.04, 64bit
I use the following command with the file:
./wavegain wavegain_memory_gain.wav
and get:
Warning: INVALID format chunk in wav header.
Trying to read anyway (may not work)...
*** buffer overflow detected ***: ./wavegain terminated
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x777e5)[0x7f090b3a57e5]
/lib/x86_64-linux-gnu/libc.so.6(__fortify_fail+0x5c)[0x7f090b44715c]
/lib/x86_64-linux-gnu/libc.so.6(+0x117160)[0x7f090b445160]
/lib/x86_64-linux-gnu/libc.so.6(__fread_chk+0x165)[0x7f090b445855]
./wavegain[0x40f797]
./wavegain[0x410ee0]
./wavegain[0x41e20d]
./wavegain[0x41c660]
./wavegain[0x40374e]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xf0)[0x7f090b34e830]
./wavegain[0x403ca9]
======= Memory map: ========
00400000-0042a000 r-xp 00000000 103:01 20397804 wavegain/wavegain
00629000-0062a000 r--p 00029000 103:01 20397804 wavegain/wavegain
0062a000-0062b000 rw-p 0002a000 103:01 20397804 wavegain/wavegain
0062b000-00668000 rw-p 00000000 00:00 0
00a1c000-00a3d000 rw-p 00000000 00:00 0 [heap]
7f090b118000-7f090b12e000 r-xp 00000000 103:01 10490464 /lib/x86_64-linux-gnu/libgcc_s.so.1
7f090b12e000-7f090b32d000 ---p 00016000 103:01 10490464 /lib/x86_64-linux-gnu/libgcc_s.so.1
7f090b32d000-7f090b32e000 rw-p 00015000 103:01 10490464 /lib/x86_64-linux-gnu/libgcc_s.so.1
7f090b32e000-7f090b4ee000 r-xp 00000000 103:01 10487655 /lib/x86_64-linux-gnu/libc-2.23.so
7f090b4ee000-7f090b6ee000 ---p 001c0000 103:01 10487655 /lib/x86_64-linux-gnu/libc-2.23.so
7f090b6ee000-7f090b6f2000 r--p 001c0000 103:01 10487655 /lib/x86_64-linux-gnu/libc-2.23.so
7f090b6f2000-7f090b6f4000 rw-p 001c4000 103:01 10487655 /lib/x86_64-linux-gnu/libc-2.23.so
7f090b6f4000-7f090b6f8000 rw-p 00000000 00:00 0
7f090b6f8000-7f090b800000 r-xp 00000000 103:01 10487658 /lib/x86_64-linux-gnu/libm-2.23.so
7f090b800000-7f090b9ff000 ---p 00108000 103:01 10487658 /lib/x86_64-linux-gnu/libm-2.23.so
7f090b9ff000-7f090ba00000 r--p 00107000 103:01 10487658 /lib/x86_64-linux-gnu/libm-2.23.so
7f090ba00000-7f090ba01000 rw-p 00108000 103:01 10487658 /lib/x86_64-linux-gnu/libm-2.23.so
7f090ba01000-7f090ba27000 r-xp 00000000 103:01 10485855 /lib/x86_64-linux-gnu/ld-2.23.so
7f090bbf5000-7f090bbf9000 rw-p 00000000 00:00 0
7f090bc25000-7f090bc26000 rw-p 00000000 00:00 0
7f090bc26000-7f090bc27000 r--p 00025000 103:01 10485855 /lib/x86_64-linux-gnu/ld-2.23.so
7f090bc27000-7f090bc28000 rw-p 00026000 103:01 10485855 /lib/x86_64-linux-gnu/ld-2.23.so
7f090bc28000-7f090bc29000 rw-p 00000000 00:00 0
7ffde0294000-7ffde02b6000 rw-p 00000000 00:00 0 [stack]
7ffde0339000-7ffde033c000 r--p 00000000 00:00 0 [vvar]
7ffde033c000-7ffde033e000 r-xp 00000000 00:00 0 [vdso]
ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0 [vsyscall]
Aborted
I use AddressSanitizer to build wavegain, this file can memory leak with the following command:
./wavegain wavegain_memory_gain.wav
This is the ASAN information:
Warning: INVALID format chunk in wav header.
Trying to read anyway (may not work)...
Warning: Unexpected EOF in reading WAV header
Unrecognized file format for wavegain_memory_gain.wav
WaveGain Processing completed normally
=================================================================
==3874==ERROR: LeakSanitizer: detected memory leaks
Direct leak of 92 byte(s) in 1 object(s) allocated from:
#0 0x7f6ca685030f in strdup (/usr/lib/x86_64-linux-gnu/libasan.so.2+0x6230f)
#1 0x4111bc in alloc_node wavegain/main.c:70
#2 0x4113d3 in add_to_list wavegain/main.c:104
#3 0x403d7e in process_argument wavegain/recurse.c:583
#4 0x4144f1 in main wavegain/main.c:718
#5 0x7f6ca613b82f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2082f)
Direct leak of 40 byte(s) in 1 object(s) allocated from:
#0 0x7f6ca6886602 in malloc (/usr/lib/x86_64-linux-gnu/libasan.so.2+0x98602)
#1 0x4097e8 in wav_open wavegain/audio.c:646
#2 0x407832 in open_audio_file wavegain/audio.c:362
#3 0x414dec in get_gain wavegain/wavegain.c:181
#4 0x41198c in process_files wavegain/main.c:197
#5 0x41453e in main wavegain/main.c:729
#6 0x7f6ca613b82f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2082f)
SUMMARY: AddressSanitizer: 132 byte(s) leaked in 2 allocation(s).
I think it would be nice if executables of version 1.3.2 could be provided as part of the release assets. Currently those assets only consist of the zip and tar.gz files, which is basically the same as selecting the 1.3.2 tag in GitHub then clicking on "Code" and then on "Download ZIP".
I understand that it might not be possible for every OS (e.g. Mac OS) but maybe for Windows and Linux executables could be added?
Hi @MestreLion .
Subj. See also Sound-Linux-More@0233391
We have a patch for gcc-10 here:
https://bugs.gentoo.org/706692
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.