A number of apps offer downloadable and/or physical scores.
They differ in various respects,
such as the source of these scores.
None of the apps has good discovery tools.
If you try, you'll wade through a lot of pieces
that you don't like, or are too easy or hard for you.
offers PDFs of public-domain and Creative Commons licensed scores:
mostly old scores that are out of copyright,
but also recent compositions by unknown composers.
Everything is uploaded by volunteers,
who typically also provide detailed metadata.
IMSLP is based on Mediawiki.
There are some limited ways to search on metadata,
but no general database search capability.
The basic search function is a Google search on the title/composer strings.
I have previously proposed
a number of ways in which IMSLP could be improved.
- A number of apps sell
commercially-published in-print scores, e.g.
Amazon, and various music publisher sites,
such as Boileau.
Few of them have significant metadata or discovery features
(although Boileau makes an effort).
has a bunch of scanned published works.
Free after 1-time $16 fee.
No search tools other than text search.
sells bound versions of out-of-print scores,
mostly by obscure 19th-century composers,
obtained (I guess) by scanning library copies.
It lets you search by composer name, and little else.
- MuseScore sells user-uploaded
scores (in an editable form, not PDFs).
These are mostly existing compositions;
there are some original compositions.
I don't think the uploaders get paid, so I'm not sure why they do it.
They have some metadata and search functions.
They collect ratings and use them to show average rating.
They also offer (for sale) some instructional materials.
let users upload scores, including their compositions.
Some of the scores are for sale.
Users can also upload MP3s or MIDI files of these scores.
It has basic metadata (style and instrumentation)
and minimal search features.
It has some social features:
rankings (averages), comments, skill level, collections.
It has user pages,
with pictures, uploads, and comments.
It's based in France.
The web site is scruffy; it's a labor of love.
The stated goal is to connect composers and performer.
They say they have 177K compositions (PDF)k,
150K MP3 files, and 34K MIDI files.
It also has an extensive directory of hundreds of
other web sites offering free scores -
largely the personal sites of composers and arrangers.
offers 2124 pieces of music, free.
They have basic metadata (instrumentation, composer, style).
Volunteers upload pieces in LilyPond format,
as Github pull requests (everything is stored on Github).
Interesting but dead.
315K scores and some audio and MIDI files.
Free; I can't figure out where they get their content.
Various attribute-based search tools,
with an "order by popularity" option for results.
The notion of difficulty
In finding a piece to play, there are two main factors:
whether you like it, and whether it's too hard or too easy for you.
Difficulty is subjective.
A piece can be easy for one performer and hard for another.
In fact, difficulty is essentially like another kind of rating,
and we can use the same techniques.
We can show the average difficulty rating, but we can also
use collaborative filtering to predict how hard
a piece will be for a particular performer.
Systems that show musical scores
should also collect difficulty ratings for compositions.
For ensemble pieces, these would be per part;
a piano trio might have a hard piano part and an easy viola part.
Difficulty ratings can be added to search interfaces in various ways, e.g.:
- Show pieces that (maximize predicted rating and)
minimize predicted difficulty.
- Define a difficulty threshold, and show
pieces just below that (i.e. you want a challenge).
- For ensemble pieces, do this separately for each performer.