AnthemScore revisited

[ted]

I have been experimenting with AnthemScore, which I bought last year but haven't used much. A long time ago I posted an improvisation, https://www.pianostreet.com/smf/index.php?topic=62945.msg672241#msg672241, recorded with my old tape system, which our late friend David April really liked. Rather than regurgitate the old thread, I post here the original mp3 together with the score and mp3 as "heard" by AnthemScore. Using the minimum quantisation of 128th notes and sticking in a lot of pedal in MuseScore produced a sound which, it seems to me, is a pretty impressive representation of the sound of my playing in the original mp3. I found at once that AnthemScore is very accurate for regularly metrical idioms such as stride and ragtime, to the extent that I probably wouldn't buy sheet music for these things again. Sure, some notes are wrongly distributed on the staff and there is the odd rhythmic fumble, but provided the original recording is halfway decent there is nothing that matters for an experienced player just wanting to learn a piece. I cannot post examples involving famous players here for obvious reasons but I can say I was impressed.

The first mp3 is my original improvisation, the second is what AnthemScore "heard" and the pdf is what MuseScore generated from it. Does this mean I would use it on my 1000 hours or more of improvisation ? No, of course not, not unless I wanted to pick up some small section somewhere. Nonetheless, I was struck by how close the sound was, given the totally unmetrical and rhapsodic nature of most of my playing.



 

[anacrusis]

Well it's transcribing well enough. I wonder if one could fine tune the AI algorithm so that it understands that you probably did not mean to play some chords 128th-note off the beat... but probably just intended to play on the beat.

With a lot of tidying up of the score I think this could become a fairly metrical composition that the player is allowed to play with some freedom.

[ted]

With a lot of tidying up of the score I think this could become a fairly metrical composition that the player is allowed to play with some freedom.

Yes that would certainly be a useful possibility for those who use improvisation as an aid to written composition. In the time it took me to do it I could record about twenty more improvisations so I wouldn’t bother myself. Either way the problem of accurately extracting pitches from a piano solo now seems pretty well solved, at least for recordings of reasonably high quality.

[lostinidlewonder]

Useful to cut and paste parts that are of interest in improvs that you want to develop, good tool. I used it for a bit too and had some fun with it, it helped me to play back improvs I did without having to solve it by ear which saved a bit of time!

[ronde_des_sylphes]

The results look similar to the aftermath of using a combination of this audio to midi


https://github.com/bytedance/piano_transcription?fbclid=IwAR0iTAav8jUYaAJw21C8eZmgdJK2Fmmgis-vSoP3lN-cOvYSYCojenoAc3Q

and putting it into musescore, except a bit tidier. The same problems in the sense that a lot of editing would be required before it's truly viable as a score. I'm not sure such things are that much better than just writing improvisations out by ear but sometimes they can be of assistance clearing up troublesome passages.

[ted]

The results look similar to the aftermath of using a combination of this audio to midi


https://github.com/bytedance/piano_transcription?fbclid=IwAR0iTAav8jUYaAJw21C8eZmgdJK2Fmmgis-vSoP3lN-cOvYSYCojenoAc3Q

and putting it into musescore, except a bit tidier. The same problems in the sense that a lot of editing would be required before it's truly viable as a score. I'm not sure such things are that much better than just writing improvisations out by ear but sometimes they can be of assistance clearing up troublesome passages.

Thanks for the link, it looks interesting. Aside from making scores of metrically regular music such as stride and ragtime, which it does close to perfectly, I probably shan't use AnthemScore that much. It also depends to what extent an improviser habitually thinks in terms of easily notated formations and rhythms to start with. Most players, through the conventional processes of musical education, seem to do that whether they are aware of it or not. Rhythm generally is a much deeper and wider internal perception than the ability of visual notation to communicate it anyway so perhaps we ought not to expect too much of these programs. 

[ted]

Useful to cut and paste parts that are of interest in improvs that you want to develop, good tool. I used it for a bit too and had some fun with it, it helped me to play back improvs I did without having to solve it by ear which saved a bit of time!

Yes, and to do that it doesn't matter if the score looks like a dog's breakfast provided all the pitches are there. It is also good for seeing at a glance what famous improvisers do in very rapid passages and, if I like the sound, storing the formations in my own vocabulary. With a Woolworths ear like mine I often imagine what is being played must be complicated and exotic but frequently find it's just an ordinary chord or scale with a couple of extra notes stuck in. I had imagined the example I posted to contain rich harmony (so did David apparently, which is interesting) but AnthemScore revealed the whole thing lies 99% within one scale ! The power of imagination and illusion cannot be underestimated !

[ranjit]

Thanks for the link, it looks interesting. Aside from making scores of metrically regular music such as stride and ragtime, which it does close to perfectly, I probably shan't use AnthemScore that much. It also depends to what extent an improviser habitually thinks in terms of easily notated formations and rhythms to start with. Most players, through the conventional processes of musical education, seem to do that whether they are aware of it or not. Rhythm generally is a much deeper and wider internal perception than the ability of visual notation to communicate it anyway so perhaps we ought not to expect too much of these programs.

You know, I completely agree with you that turn rhythm is much deeper than the confines within which a majority of people operate. But how do you convince those people?! I suppose you don't, but I would very much like to.

[ted]

You know, I completely agree with you that turn rhythm is much deeper than the confines within which a majority of people operate. But how do you convince those people?! I suppose you don't, but I would very much like to.

No, I don't, I'm too much of an expert at being wrong to try (ask my wife !) Rhythm is a subjective phenomenon along the continuous time-line while notes, harmony, chords and the rest are wholly discrete entities, at least for fixed pitch instruments. Therefore the latter can be objectively studied, made academic, have theories made about them, while the former is much more difficult to understand. How even a simple combination of notes is played, arpeggiated, phrased, renders it a completely different musical entity. To assert that only the combination of notes carries musical meaning seems to me destructively simplistic. That's just how I see things, thousands would disagree.