If you want to create music with free software on Linux, you can choose from a few digital audio workstations (DAWs). If you play live music, Ardour is usually a good choice. However, if you primarily compose music in MIDI notation with virtual or hardware synthesizers, you may want to consider MusE 4. Of the free DAWs for Linux, MusE 4 puts the most emphasis on full support for the many methods and standards that have found their way into MIDI technology over the past 50 years.

Where to Get It

A few years ago, MusE was in the standard feature set of popular distributions related to music production, which is true of Ubuntu Studio (used to test MusE in our lab). Today, the major distributions often only have an outdated version in their package sources because work on MusE has been pretty slow at times.

To get the latest version, you can download an AppImage from the MusE website [1]. AppImages (and containers like them) are not necessarily the perfect solution for real-time audio. The elaborate wrapper increases the system load and the compartmentalized image often also prevents correct integration with the JACK audio server and ASLA MIDI. However, the MusE team has done a good job of testing the AppImages very carefully. A tolerable delay at startup time is the only side effect of running MusE from the AppImage. Once Muse4 is running, it responds quickly and integrates into the environment in an exemplary manner.

You also have the option of building MusE from source code. If you go this route, you will need a complete, up-to-date build environment for Qt5, as well as about a 100 other audio development packages. If you have built Ardour, these packages will already be in place on your system. However, because the AppImage contains the current release of the source code and works very well, the overhead of compiling the source code is only worthwhile in exceptional cases.

After downloading, the AppImage only needs the right permissions to run; when it has them, the program starts up nicely (Figure 1). In our lab, it detected a JACK server that was already running, without any manual intervention.

Figure 1: After making the AppImage file executable, you can launch MusE immediately.

MusE does not provide the pre-made loops and examples you may be familiar with from commercial software. Instead, MusE supports everything installed on the system, including plugins in the native Linux VSTx format. If you want to use VSTs available as Windows DLL files, you will need a converter such as yabridge [2] (Figure 2). On Ubuntu 20.04 LTS, you will need a more recent version of Wine, which you can download from the Ubuntu Wine repository [3].

Figure 2: Thanks to yabridge, the Feldspar freeware software synth in the form of a Windows DLL works flawlessly in MusE.

Peculiarities

MusE follows the same principles as other DAWs: It provides an automatable tape recorder plus a signal mixer. However, the way MusE implements these principles requires some getting used to for Ardour or Bitwig users.

At startup, MuseE offers to initialize external and virtual MIDI sound generators. You can send setup commands to these devices to configure them individually, which shows that MusE isolates the sound generators from the actual music data more than other DAWs do. As your first step, you need to set up the sound generators and the physical audio inputs and outputs. You can then create tracks that feed these devices with MIDI signals and audio data (Figure 3).

Figure 3: MusE's routing tool lets you wire signal sources (left) and sound generators and audio outputs (right). You load plugins individually to access them from the list on the right.

The configuration tool for the MIDI port connections (Figure 4) turns out to be anything but intuitive (which might cause some people to quickly give up on the program). While the configuration tool displays all possibilities, it does not let you connect the detected MIDI devices to anything in the project – despite a tooltip prompting you to do exactly that. The tooltip fails to point out that you first need to set up the devices as ports for MusE before the program offers them as inputs for MIDI tracks (see the "Setting Up a Studio" box).

Figure 4: In the MIDI Port and Soft Synth Configuration tool, a list located bottom right lets you map the existing devices to ports for the tracks in MusE.

Setting Up a Studio

Before you record MIDI tracks and send them to the sound generators, you need to set up a device in MusE for each generator. The device will appear as a track in the Arranger. MusE gives you two options for setting up a device – they differ greatly in some respects. The first option lets you insert synth tracks in the Arranger simply by right-clicking. When you do this, MusE displays the list of plugins that are available and supported by the system. After selecting the desired plugin from the list, you can then use it as an instrument. In the background, MusE creates a new entry in the list, which you will find in Edit | Midi Ports | Soft Synths.

However, the first option prevents you from using the same plugin more than once. This would be a very strange restriction if you could not use multiple instances of the same plugin. Especially with samplers, it's quite common to have half a dozen individual noise spinners. From a purely technical point of view, this is not a problem in all of the current plugin systems.

You can get around this restriction with the second option. If you create the synth directly in the list below Midi Ports | Soft Synths, the restriction disappears. Now when you create a new instance, it also appears immediately as a Synth track in the Arranger (Figure 5).

Figure 5: If you double-click on the KickMaker plugin entry in the list, it will be added to the project and automatically preconfigured correctly for use as a MIDI note target.

For editing details in the recorded material, MusE offers its own editors; you can open an editor in the tabs by double-clicking on a recorded or imported part (a chunk of coherent notes or wave data in MusE terminology). The main Arranger view always stays open as the first tab bottom left.

MusE removes the zoom limit in the Arranger view. Depending on the task, you will see various new toolbars and functions. MusE's wave editor, however, turns out to be very unwieldy and not very intuitive (Figure 6). After some familiarization, the wave editor can be used to perform precise cutting and simple actions such as normalizing. However, for anything beyond that, the wave editor's design is virtually incomprehensible, and the tool sometimes crashes (see the box "Wave Editor Issues").

Figure 6: Selecting Wave edit from the context menu (right-click) lets you call up the wave editor, which is somewhat awkward to use. However, you can find important functions such as normalizing in the Functions menu top left.

In comparison to the wave editor, the MIDI editor tabs are state-of-art and contain a multitude of useful functions and tools, as you might expect. The only downside is the graphical controller view at the bottom: It looks intuitive and practical, but the pencil tool only lets you edit the controller values for all notes. The tool does not take into account the selection of notes at the top of the canvas. This proves to be particularly annoying for drum notes because it is not possible to individually set the beat for a snare drum if it is in the same bar as a bass drum. While this type of important operation is not impossible in MusE, it is unnecessarily awkward.

Some important functions are missing in the menus, but you can access them via keyboard shortcuts. For example, Ctrl+B lets you insert MIDI parts as clones (Figure 7). Changes to the original have the same effect on the clones. Ctrl+V lets you create independent copies for individual editing. If you are working with bass/drum loops, working with clones is far more efficient. A note that is not quite perfect in a part will be in the right place in the clones after correcting it in the editor.

Figure 7: MusE shows clones as a dashed border instead of a solid one. Press Ctrl+B to turn the currently selected part into a clone.

Finally, MusE's score editor offers a unique selling point (Figure 8). Even the expensive Bitwig Studio does not offer a classic score view, let alone a score editor. A popular alternative on Linux, MuseScore [4] (which, as the name suggests, comes from the MusE project) offers even more functionality for classical composition (see the box "History").

Figure 8: A part for a monophonic synthesizer in the score editor: Classic techniques such as legato are far easier to implement with the score editor than in the simple bar matrix.

Many Tweaks

Many DAW functions can already be controlled directly in the MusE interface. To compensate for the rather sparsely populated main menus, however, very rich context menus pop up everywhere. The few tools that you open via the main menus offer a variety of additional functions in their own dialogs.

To save space, the function table at the top of the Arranger on the left only expands by one third in the default setting. You will find entries here that offer many functions you may have missed at first glance. One of them, for example, lets you call up the plugin's graphical interface as well automate all of the parameters in synth tracks. The ability to automate the parameters in synth tracks clarifies the purpose of the track for a sound generator because it does not contain any music data in MusE (Figure 9).

Figure 9: The automation track for the complex Calf Monosynth sits in its own plugin track, instead of the MIDI tracks controlling it. You can control anything that the synth's interface lets you control here.

Assigning the automation curves directly to the synth makes them obsolete in the recorded tracks. If you want a different kind of automation for another recorded track, you have to create an individual plugin instance for it. If you are familiar with the playlist concept in Ardour, you will already know about this concept.

Score tracks in MusE basically work like playlist tracks in other applications: one sound generator, multiple data tracks. Somewhat confusingly, the data tracks can contain individual automations, such as those you apply with the keyboard's modwheel during recording. This is conceptually consistent because these automations do not belong to an individual synthesizer but are generic MIDI parameters, such as vibrato or volume, that are the same for each instrument.

MusE comes with dozens of presets for popular classic synths and many for modern plugins. When you select a preset for your instrument, the program embeds the control signals for the instruments into the data tracks. In addition, individual initializations can be implemented as scripts. There are no special tools for this, and it requires a good deal of prior knowledge, but if you know your synths and their interfaces inside out, you can do practically anything with them. This also applies to hardware synths connected via the MIDI sequencer in JACK or ALSA.

For newcomers, it makes more sense to rely on MusE's sensible automations and ignore the complexity. The usual synth plugins such as Yoshimi or Calf Monosynth work flawlessly even if you don't manually tweak the settings. The same applies to commercial Linux VST modules such as Mackie's complex BioTek synth. BioTek's registration lets you use it in MusE even if you bought BioTek in a bundle with Mackie's Waveform (Figure 10).

Figure 10: After registering with your email address and a password for your Mackie account on the Tracktion website, BioTek works flawlessly in MusE.