I have composed several tracks using MuseScore. I exported these as MIDIs and played them on my phone to my friend. His comment was: "It sounds like the music from an old computer game like Super Mario".

This was not exactly the vibe I was going for, as I was hoping these would sound a little more like real instruments (or at the least not so tinny and obviously electronic). I tried exporting in different formats, eg MP3 but the sound was not noticeably different to my friend.

I tried searching for ways to improve the sound quality however a lot of these were not practical for me, such as "adjust the velocity of each note in the composition" and "record yourself playing the real instruments".

Therefore I wanted to ask, does anyone know of any simple methods I could use to improve the sound quality? I do not mind buying a program if this will improve the quality significantly. I appreciate anyone's time and thank you in advance.

  • 2
    The problem are the sound samples. The libraries you use are low quality – Shevliaskovic Jun 6 '19 at 14:08
  • Unfortunately, the problem is not the composition or the software, it's the performance, or lack thereof. The performance of a composition is critical to bringing out the feeling of the composition. If you're composing EDM, then having a mechanical/electronic sound works, but for almost every other genre, the only way to make it sound right is to either have it actually performed (and recorded) or to make adjustments to the sequenced data to make it emulate a performance. – Todd Wilcox Jun 6 '19 at 14:53
  • Granted, even MP3s and WAVs using Musescore's default soundfonts sound substantially better to me than MIDIs created with Musescore. – Dekkadeci Jun 6 '19 at 15:59
  • What version of Musescore are you using? – Nevin Williams Jun 6 '19 at 23:26

An exported MIDI file will play using the default soundset of the device that's playing it. It's amazing that Windows, Android, iOS etc. even HAVE a built-in MIDI playback system with reasonably recognisable instrument sounds! But yes, it's a bit basic.

A more interesting question is what can be exported as audio from MuseScore and shared as a MP3. MuseScore uses SoundFonts for its internal playback. It's rather old technology. Alternatives to the default sound set are available, as detailed in the MuseScore manual.


One of the advantages of using the more advanced (and much more expensive) score publishing programs (Finale, Sibelius, Dorico...) is the wider choice of sound sets, and more developed systems of making them sound 'real'. NotePerformer, available for the high-end scoring programs, is particularly ingenious in making lyrical music styles play effectively. But the best realism currently available is probably from transferring your score into a sequencer program (Cubase, ProTools etc.) where you'll have access to the best available sample sets and meticulous control over their use.

Join a forum dedicated to Sibelius, Dorico or Finale. Ask if someone will take your .mscz file from Musescore, run it through the various playback systems he has available and return it to you as a MP3. People are often remarkably happy to be helpful, particularly when it involves showing off a piece of software that they've invested heavily in :-)

(If you want me to do this for you, Googling my name will find my website very easily, contact email is at the top of the homepage.)

  • 1
    I think this is placing too much emphasis on the samples and not enough on the performance and/or performance humanization algorithms. It's the humanization that makes the biggest difference, in my experience. Also note that Sibelius can be rented on a monthly basis for a reasonable fee for anyone looking to have an affordable way to humanize a few compositions and then stopping the subscription. – Todd Wilcox Jun 6 '19 at 14:59
  • Sibelius, in itself, won't do much 'Humanising'. I'm not sure if the free trial of Sibelius can be combined with the free trial of NotePerformer? But remember that NP, while effective in lyrical styles, tends to sabotage 'groove' based music with it's 'expressive' alterations of rhythm. A feature that isn't optional. It would be good to see a sample of @Appguy1's music. Then we could offer informed suggestions on how it might be improved. – Laurence Payne Jun 6 '19 at 15:20
  • Having been a recent subscriber to Sibelius and a current owner of Finale, I can say with great confidence that they both have reasonably good humanization algorithms. – Todd Wilcox Jun 6 '19 at 15:30

The simplest (but not necessarily most effective) way to improve the performance quality is to either export MusicXML or MIDI from MuseScore and open it up in a different software package that can either automatically add humanization or performance data or a package that makes it easy to add and adjust humanization.

As Laurence noted, if you export to MusicXML and then import to Finale, Sibelius, or Dorico, each of those three will create a much more human performance from the same MusicXML data compared to MuseScore. MuseScore has very little in the way of performance humanization.

Alternatively, and more powerfully, you could export MIDI and import the MIDI into a full-featured DAW such as Cubase, Logic, or Pro Tools. All three of those packages include mass MIDI editing tools that you can use for adding randomness to both timing and velocity. Other DAWs may have similar tools and capabilities, but I'm not sure. Randomization is only going to get you a small benefit. It won't sound as good as the humanization of one of the above programs.

The second-best-sounding approach is to use a DAW and make a long series of tweaks to the MIDI data, using what you know about the actual instruments, and also taking advantage of the tools in better sample libraries. For example, a lot of string sample libraries will have different samples for different techniques, like pizzicato, tremolo, etc. So you would go through and make sure you program in the sample switching for each technique. For regular techniques, you also want to make sure there are performance tweaks, such as selecting every other note of eighth note or sixteenth note runs and sliding the timing slightly and bumping the velocity to imitate left hand/right hand switching or upstroke/downstroke or whatever is appropriate for the instrument. I know you don't want to spend all the time to do this, but without this level of effort the results will also be mediocre at best.

The best sounding approach is to hire musicians to perform it for real. Some instruments, such as percussion, can sound very good with sound samples and slightly humanized playback, so you could hire only a couple musicians for the more sensitive instruments. If you have a whole string section, you could just hire one violinist (say) and have them record several takes and use that and some delay and chorus effects to create a much larger sound.


One way I've found that makes a big improvement to MP3 exports of piano pieces is by bumping up the reverb a notch. This can be done under View...Synthesizer, selecting the Master Effects tab, and click-dragging the right-most knob Output from its default position, ~25% Dry, to an even mix (as shown below), or even a bit wet.

Reverb Controls

If your composition has several instrument tracks, it can also be helpful to set their individual left/right panning, in View..Mixer, or F10 to give the instruments some spatial separation.

A search on "Mixing a digital orchestra" brings up a few articles by Mattias Westlund that describe how to roughly pan a digital orchestra and why one should.

Here is an excerpt from that page:

1st violins – halfway left
2nd violins – less than halfway left 
    (close to 1st, but you should be able to tell them apart)
Violas – center or slightly right
Celli – less than halfway right
Basses – halfway right
Trumpets – one third right
Horns – one third left
Trombones/tuba – halfway right
Flutes/clarinets – slightly left
Oboes/bassoons – slightly right
Percussion – timpani and bass drum centered or to the left 
    (so as not to clash with celli/basses), anything else to taste

Arn Andersson has an article on this topic, with examples you can listen to, and well explained step-by-step instructions:


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.