Gigsplit

GigSplit: Splits long mp3 and multitrack recordings

This is a web app for splitting a long mp3 recording into separate songs. It can also split a folder full of wavs files, one per channel, and produce a folder full of multitrack songs ready for mixing.

Written using Svelte, peaks.js, and jExcel. Works under Linux, but could probably be made to work elsewhere if the command-line audio tools can be made available.

Requirements

Requires node and npm to be installed.

The following audio tools must be available on your server (in your path):

1. audiowaveform https://github.com/bbc/audiowaveform
2. sox sudo apt install sox
3. cutmp3 sudo apt install cutmp3
4. id3tool sudo apt install id3tool
5. flac sudo apt install flac

Installation

1. Clone this repository and cd there
2. npm install
3. ./copyCss.sh
4. Other patches that might be necessary to eliminate warnings are listed in ./patches.md
5. npm run dev or npm run build and npm run start

Usage

GigSplit requires a data directory containing an mp3 file and, optionally, a directory full of individual channel recordings in wav format. Both are identified by the date of the recording in the format YYYYMMDD. For example, for a recording on 12/05/2019, the data directory should contain 20191205.mp3 and a directory of wavs called 20191205.wavs/ (there may be other files in the directory, they will be ignored).

The data directory should be symlinked (or copied, if you want) into ./static with the name data. I usually have a separate directory for each recording, but because the files are identified by the date in the file names, you could process multiple recordings in the same directory. The mp3 files for the individual songs will be written to ./static/data/<band name in lowercase>-<date> and the multitrack splits for each song will be written to ./static/data/multitrack/<date>-<song title>/. The files in the multitrack directory will have the same names as the original wav files and will be compressed to flac format.

Go to localhost:3000 in your browser. On the Settings page, enter the band name, album name, date, and genre. On the Split page, it should find your mp3 file and prompt you to create the corresponding .dat file (containing the peaks for the waveform display). If not, go back to the Settings page and fix the date, and verify that ./static/data contains your mp3 file and wavs directory.

Once the waveform is shown on the Settings page, you can begin splitting the recording. The middle button below the waveform shows the current position and starts/stops playback. The buttons to the right and left advance and retreat. Once you've selected a line in the table, the buttons above the table allow you to copy the current time (now) to the start or end column, or copy the values in the table to now.

Once the table is full go to the Processing page, where you can create the directories, split the mp3 file, and split the wavs. After each mp3 song is created GigSplit will tag the file with the info you provided on the Settings page. After splitting the wav files GigSplit will compress them, losslessly, to flac and remove the wav files. Your original mp3 and wav files are never modified.

Typical Workflow

1. Record your session in multitrack. The free version of Tracktion works fine for this, but you can use any DAW software that records to wav. Don't worry about pausing the recording, since you'll be cutting it later. I only pause when we take a break. Use a Tracktion preset that has all of your channels named with the instrument and performer (this will end up in the wav filenames). Save your Tracktion session.
2. If you never paused while recording, you'll have one wav file per channel. But if you did, you can't just concatenate the files, you need to do this in Tracktion. The start times for each channel will vary slightly, and this info is in your Tracktion session. Render the session, selecting the option to generate one file per channel. I also enable normalization here, and I don't add any effects.
3. Move the wav files to your <date>.wavs directory. If you had to merge them in your DAW, then they're probably all stereo now. You might want to make the channels that were originally mono mono again, you can do this with sox <input wav> <output wav> remix 1,2. You should also remove any spaces in the filenames.
4. Create a new Tracktion project and import the files from your <date>.wav directory, one per channel. Create a rough mix, keeping in mind that you'll do the real mixes with the split files. For now, you just want to make something listenable. I add effects to a few tracks and look for the loudest sections for each channel to set reasonable levels. Render to a stereo wav file and convert that to mp3 with an external program (for example, lame -b 256 <wav file>). Move the mp3 file to <date>.mp3 in your data directory.
5. Symlink your data directory to ./static/data in your GigSplit directory, start up GigSplit and split the session.
6. Now that I have the starts and ends of all the songs, I reopen my Tracktion project and tweak the mix. I do this because, for many of the songs, no one will ever get around to mixing them. Mix down to stereo mp3, replace your <date>.mp3 file, and re-split in GigSplit.
7. I also normalize the gain of all of the mp3 songs using mp3gain -r -c *.mp3 in the output directory.
8. Burn copies of the multitrack wav files, give them to your bandmates, and tell them to mix them. You've done enough work already!