DJ Mix Master

Masters recorded DJ mixes. Open a long recording, mark the mix inside it, and the tool evens out the loudness from track to track: louder tracks are turned down and quieter ones up, while breakdowns keep their natural dip instead of being pumped up.

It comes as a desktop app for macOS, Windows and Linux, and as a command-line tool that shares the same processing. It is free, and will stay free; if it helps you, you can sponsor it.

Download and install

Installers for macOS (Apple Silicon), Windows and Linux are on the website and the releases page; on Windows it is also in the Microsoft Store. Nothing else needs to be installed; the app carries everything it uses. The installers are not signed, so each system warns about them the first time; the manual's Installing page says how to get past that. What changed in each release is in CHANGELOG.md.

There is no build for Intel Macs yet. On those, the tool can be run from source.

The manual

The manual covers every part of the app, with pictures: opening a recording and marking its mixes, the check-up, levelling, tracks, exporting, and the videos, down to every kind of layer and every function a template's formulas can use. Its pages are in site/manual/; the published one describes the latest release. The app carries its own copy, for its own version, and Help → User Manual opens it in a window of its own, with a search, without a network.

Getting started

1. Open a recording. Drop a file on the window or use "Open recording…". It is analysed once. A recording split into several files by the recorder opens as one when its parts are chosen together; a video of the set opens like a recording.
2. Mark the mix. The whole recording starts out as one mix. Drag its edges in, or press "Split at gaps" to cut a night apart at the silences.
3. Listen. Press Space to play, and A to hear the recording as it was.
4. Adjust if needed. The defaults work for most mixes.
5. Export the sound, or a video of the mix moving to the music.

Your recording is never changed and never leaves your computer. The only thing the app contacts is GitHub, at most once a day, to see whether a newer version exists. If there is one, a line in the header says so; nothing is installed for you. Turn the check off in Settings.

What it does

• Levels track against track, not moment against moment. A slow gain curve follows how loud each track is at full energy. Quiet passages shorter than about 90 seconds are treated as breakdowns and keep the gain of the track around them.
• Optional bass correction. Kick and bass can be levelled separately from everything above them with a slowly moving low-shelf EQ, so the low end stays consistent without dragging the mids and highs along.
• A corrective EQ per mix, for a phono stage that sounds dull or harsh, a PA mixer's EQ or turntable rumble: a rumble filter and four broad bands, applied before the levelling, which measures the mix as it sounds with it.
• Leaves headroom by default. With no target set, the result is made as loud as fits under the ceiling without leaning on the limiter. A higher target pushes into the limiter for more level.
• Look-ahead true-peak limiter on the output, with the ceiling at -1 dBTP by default. It holds the level between samples as well as on them.
• A check-up of the recording: clipping, channel faults (with one-click fixes), dropouts, hum and more.
• Tracks and chapters: mark where tracks start, import a list from DJ software, copy a list for YouTube; exports carry them as chapters.
• Videos for YouTube and social media, in landscape, vertical or square: templates that move to the music, or the recording's own picture with the mastered sound. Templates can be made in an editor with layers, effects, blends, layer styles and formulas, and shared as files.
• Records the night itself from any two inputs of a DJ mixer's sound card or any other input, with its level shown before you start, in parts of half an hour by a program of its own that carries on if the app closes, and opens the recording when it stops.
• Handles long files. Recordings are streamed, never loaded whole.

Limitations

• Clipping cannot be undone. A recording made too hot is turned down to match the rest, but tracks that were clipped still sound clipped.
• Long quiet passages. A passage quieter than the rest for longer than the breakdown length counts as a quieter track and is raised, up to the boost limit.
• Handovers without a gap are not found. "Split at gaps" sees silences only. Where one DJ mixed out of another, the split is yours to place.
• A recording that was edited or converted is a new recording to the app: its saved mixes stay with the old file. Moving or renaming it is fine.
• Less tested on Windows. It is developed on macOS. The Windows build has been used on a real machine but gets less testing.

The command line

The same processing without the window. It needs Node 24 or newer and ffmpeg installed; see Running from source.

node src/cli.ts recording.mp3 --start 1:12:00 --end 2:45:30 -o mix.wav

The output format follows the file extension: .wav, .flac, .mp3 or .m4a. Without -o the result is written next to the input as <name>.mastered.wav.

Option	What it does	Default
--start, --end	The mix inside the recording, in seconds or h:mm:ss	whole file
-o, --output	Output file	<input>.mastered.wav
--bit-depth <16|24>	Bits per sample for .wav and .flac; 16-bit is dithered	16
--bitrate <kbps>	Bitrate for .mp3 and .m4a	320 and 256
--sample-rate <Hz>	Sample rate of the output	the input's
--fade-in, --fade-out <seconds>	Fade at the start and at the end	none
--headroom <dB>	Extra room under the ceiling for .mp3 and .m4a	1
--title, --artist, --album	Tags to write into the output	none
--force	Overwrite the output if it exists	off
--analyze-only	Print the planned gain without writing audio	off
--dump <file.csv>	Write the loudness and gain curves for inspection	off
--target <LUFS>	Loudness of the full-energy sections	as loud as fits without limiting
--no-levelling	Leave the tracks as they are; turn the whole mix up or down by one amount	levelling on
--gain <dB>	That amount, set by hand (implies --no-levelling)	as loud as fits
--window <seconds>	Quiet passages shorter than about this are breakdowns	100
--percentile <0-1>	Loudness percentile that counts as full energy	0.9
--bass-correction <0-1>

The command line looks for ffmpeg on PATH and in the usual Homebrew folders. Set DJMASTER_FFMPEG to use a specific binary.

How it works

1. Measure. One pass over the audio produces the loudness (ITU-R BS.1770, K-weighted) in 100 ms blocks, plus the level of the low band and the peaks.
2. Find the full-energy level. A sliding high percentile fills in every dip shorter than the window, then the mirrored low percentile puts the edges back where they were. Breakdowns vanish from this curve; a change of level between two tracks stays a step.
3. Derive the gain. The gain is the difference between that curve and its typical level, limited by the boost and cut settings, smoothed and capped in how fast it may move. Hand adjustments are added on top.
4. Render. A second pass applies the bass shelf and the gain, runs the limiter and encodes the result.

Running from source

Needs Node 24 or newer and ffmpeg on the machine (brew install ffmpeg on macOS).

git clone https://github.com/Suva/djmixmaster.git
cd djmixmaster
npm install
npm start

Building the installers

npm run package

This builds an installer for the system it runs on into release/: a .dmg on macOS, a setup .exe on Windows.

To publish a version, run npm version minor (or patch) and then git push --follow-tags. That updates the version number, tags it, and makes GitHub build both installers and publish them as a release (see .github/workflows/release.yml).

Development

npm test          # build, then unit tests and tests that drive the real window
npm run typecheck

AGENTS.md explains how the code is organised and how to check changes; TASKS.md lists what could be done next.

Supporting it

DJ Mix Master is free, and it always will be: every feature, for everyone, with no accounts, no advertising and no paid version. It is made in spare time, by one person. If it saves you hours on your recordings and you would like to say thanks, you can sponsor its development on GitHub, once or every month. It helps pay for the time, and for testing on more sound cards and systems. Telling others about it, and reporting what does not work on the issues page, helps as much.

Licence

Copyright (C) 2026 Jaan Pullerits

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. It is distributed in the hope that it will be useful, but without any warranty. See LICENSE for the full text.

The installers include a copy of FFmpeg, which is licensed separately under the GPL. It is the build from the ffmpeg-static project, which documents its sources; its licence is in the app's resources folder next to the ffmpeg binary.

The fonts for videos in assets/fonts come from Google Fonts and are each licensed under the SIL Open Font License, whose text is beside each font.

The bypass icon is bypass_alt from Brightlayer UI Icons by Eaton, under the BSD 3-Clause License; its text is in assets/icons and travels with the app.