How to I go from .flac to .mp3 using LAME & FLAC using the Terminal alone?
For a long time I've been using a relative clunky technique which involves Audacity with a LAME plugin. This is fine I guess, but the appeal of the Terminal approach is I can be a little finer grained with my [options]
and perhaps use more up-to-date binaries.
Furthermore, my MacBook is ageing a little now and if I can get rid of an unnecessary GUI, all the better.
Thanks in advance.
Solution 1:
Converting a single file without preserving tags
brew install lame
flac --decode --stdout test.flac | lame --preset extreme - test.mp3
-
--decode --stdout
=-dc
-
lame - $outfile
= input from STDIN -
--preset extreme
= ~245 kbit/s VBR
A shell script that preserves some ID3 tags
#!/bin/bash
for f in "$@"; do
[[ "$f" != *.flac ]] && continue
album="$(metaflac --show-tag=album "$f" | sed 's/[^=]*=//')"
artist="$(metaflac --show-tag=artist "$f" | sed 's/[^=]*=//')"
date="$(metaflac --show-tag=date "$f" | sed 's/[^=]*=//')"
title="$(metaflac --show-tag=title "$f" | sed 's/[^=]*=//')"
year="$(metaflac --show-tag=date "$f" | sed 's/[^=]*=//')"
genre="$(metaflac --show-tag=genre "$f" | sed 's/[^=]*=//')"
tracknumber="$(metaflac --show-tag=tracknumber "$f" | sed 's/[^=]*=//')"
flac --decode --stdout "$f" | lame --preset extreme --add-id3v2 --tt "$title" --ta "$artist" --tl "$album" --ty "$year" --tn "$tracknumber" --tg "$genre" - "${f%.flac}.mp3"
done
To use the script, just save it somewhere like ~/bin/flac2mp3
and make it executable with chmod +x ~/bin/flac2mp3
.
This would convert all flac files in your Music folder:
find ~/Music/ -name '*.flac' -exec ~/bin/flac2mp3 {} \;
Or slightly faster, since it only calls flac2mp3 once:
find ~/Music/ -name '*.flac' -print0 | xargs -0 ~/bin/flac2mp3
Solution 2:
ffmpeg would preserve tags (but not cover art) by default.
for f in *.flac; do ffmpeg -i "$f" -aq 1 "${f%flac}mp3"; done
-aq 1
corresponds to -V 1
in lame. -acodec libfaac
would convert the files to AAC:
for f in *.flac; do ffmpeg -i "$f" -acodec libfaac -aq 200 "${f%flac}m4a"; done
Solution 3:
i took what you guys had, but then made it run even faster by using xargs
to parallelize the jobs.
find <directory> -name '*.flac' -print0 | xargs -0 -P8 -n1 /usr/local/bin/flac2mp3
Then this is the script from above /usr/local/bin/flac2mp3
#!/usr/bin/env bash
for f in "$@"; do
[[ "$f" != *.flac ]] && continue
album="$(metaflac --show-tag=album "$f" | sed 's/[^=]*=//')"
artist="$(metaflac --show-tag=artist "$f" | sed 's/[^=]*=//')"
date="$(metaflac --show-tag=date "$f" | sed 's/[^=]*=//')"
title="$(metaflac --show-tag=title "$f" | sed 's/[^=]*=//')"
year="$(metaflac --show-tag=date "$f" | sed 's/[^=]*=//')"
genre="$(metaflac --show-tag=genre "$f" | sed 's/[^=]*=//')"
tracknumber="$(metaflac --show-tag=tracknumber "$f" | sed 's/[^=]*=//')"
flac --decode --stdout "$f" \
| lame --preset extreme \
--add-id3v2 \
--tt "$title" \
--ta "$artist" \
--tl "$album" \
--ty "$year" \
--tn "$tracknumber" \
--tg "$genre" \
- "${f%.flac}.mp3"
done
and heres some stats for the performance speedup using parallelism.
find <dirOfFlac24s> -name '*.flac -print0 | xargs -0 -P8 -n1 /usr/local/bin/flac2mp320
0.00s user 0.00s system 60% cpu 0.002 total
115.94s user 1.40s system 359% cpu 32.655 total
time /usr/local/bin/flac2mp320 <dirOfFlac24s>/*.flac
96.63s user 1.46s system 109% cpu 1:29.98 total
you can see it also utilized my CPUs more effectively, i have an intel i7, so 8 is probably the right number of processes.
Solution 4:
Found this thread while trying to do direct encoding of MP3s from FLAC source files. Boehj’s answer provides a decent scripting option, but I personally prefer to use FFmpeg, so this is the Bash script I came up with to handle this task. Tested and works great in macOS Sierra (10.12.2).
Perquisites: You should have ffmpeg
and lame
already installed on your Mac. The easiest way to do this is via Homebrew. First make sure you have Homebrew installed like this:
/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
Then run this command to install ffmpeg
and lame
:
brew install ffmpeg lame
Once that is done you are ready to run this script. This script will look for FLAC files in the directory path/to/FLAC/files
but that can be changed to simply be .
if the FLAC files are in the same directory you are running this script in. When it runs it will create an mp3/
subdirectory where all of the MP3 files will be placed.
find -E "path/to/FLAC/files" -type f -iregex ".*\.(FLAC)$" |\
while read full_audio_filepath
do
# Break up the full audio filepath stuff into different directory and filename components.
audio_dirname=$(dirname "${full_audio_filepath}");
audio_basename=$(basename "${full_audio_filepath}");
audio_filename="${audio_basename%.*}";
# audio_extension="${audio_basename##*.}";
# Set the MP3
mp3_dirpath="${audio_dirname}/mp3";
mp3_filepath="${mp3_dirpath}/${audio_filename}.mp3";
# Create the child MP3 directory.
mkdir -p "${mp3_dirpath}";
# Get the track metadata.
mp3_title=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:TITLE= | cut -d '=' -f 2- );
mp3_artist=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:ARTIST= | cut -d '=' -f 2- );
mp3_album=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:ALBUM= | cut -d '=' -f 2- );
mp3_year=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:YEAR= | cut -d '=' -f 2- );
mp3_track=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:TRACK= | cut -d '=' -f 2- | sed 's/^0*//' );
mp3_tracktotal=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:TRACKTOTAL= | cut -d '=' -f 2- | sed 's/^0*//' );
mp3_genre=$(ffprobe 2> /dev/null -show_format "${full_audio_filepath}" | grep -i TAG:GENRE= | cut -d '=' -f 2- );
# Where the magic happens.
ffmpeg -y -v quiet -nostdin -i "${full_audio_filepath}" -ar 44100 -sample_fmt s16 -ac 2 -f s16le -acodec pcm_s16le - | \
lame --quiet --add-id3v2 --pad-id3v2 --tt "${mp3_title}" --ta "${mp3_artist}" --tl "${mp3_album}" --tn "${mp3_track}"/"${mp3_tracktotal}" --tg "${mp3_genre}" -r -m s --lowpass 19.7 -V 3 --vbr-new -q 0 -b 96 --scale 0.99 --athaa-sensitivity 1 - "${mp3_filepath}";
done
Some notes on things I learned “The Hard Way™” so others can gain from what I did differently in this script compared to others on the Internet.
- The
grep
commands for tag parsing (using FFprobe which is installed with FFmpeg) are case insensitive using the-i
option to make itgrep -i
. - The following
cut
command is now limited to dividing the output only based on the first=
in a tag name with the-f 2-
option which makes the commandcut -d '=' -f 2-
. For example, Pavement has a song titled “5-4=Unity” and if only the second chunk were selected via cut that title would have been truncated to “5-4”. - For track—and total track—numbers I added an extra pipe to
sed
which gets rid of leading zeros:sed 's/^0*//'
. - In similar scripts around the Internet, the FFmpeg output is something like
-f wav
and that would actually compress the FFmpeg output which makes no sense in a pipe setup where LAME is going to re-encode it. Instead the output here is set to-f s16le -acodec pcm_s16le
which is basically RAW output; perfect for piping audio to another process like this. - To deal with RAW output on the LAME side of the pipe, I had to add the
-r
option. - Also note the
--tt
,--ta
,--tl
,--tn
and--tg
ID3v2 tag options for LAME. When audio is streamed/piped from one process into LAME the the metadata from the source file is lost. One suggested option is to get FFmpeg to save the metadata to a text file by setting the option with-f ffmetadata "[metadata filename here]"
and then running FFmpeg again with the something like this:-i "[metadata filename here]" -map_metadata 1 -c:a copy [destination mp3 file] id3v2_version 3 -write_id3v1 1
. That works, but note the requirement for a destination file. Seems like FFmpeg only imports metadata when it can copy the file which seems like a very wasteful process. Using FFprobe to get values and then setting them in LAME with--tt
,--ta
,--tl
,--tn
and--tg
options works better; all the metadata is written in place so duplicate file needs to be generated.