It would indeed be in everybody’s interests if all the creators of invalid musicXML files get this straightened out.
But the commercial reality is that if one notation app (i.e. Finale) is prepared to invest a lot of resources cleaning up everybody else’s mess, they can just turn round and say “not our problem, it works fine with Finale”.
And being cynical, Finale needs to have some unique selling point if wants to have a future
But he didn’t offer any indication of whether it actually works very well with Dorico. I have no particular allegiance to any of the scanning suppliers. Few of them seem to be making any real effort to make their products useful in the real world. Their claims border on false advertising, IMHO.
I’d be perfectly happy if Steinberg chose one product – I don’t care which one – and invested the effort to get that one to work well end to end. I’d be happy to pay an additional price for that capability. If PlayScore is actively developing, perhaps they are the best supplier for Steinberg to work with. But given how poorly all the other products do with regard to MusicXML, I’m not eager to jump on board without a real commitment from Steinberg.
A little update : Photoscore is now communicating that the update for Catalina will come in December… Not sure how reliable this team is, it was supposed to be there in October and there is no other information apart from that little sentence in their old site.
I use PlayScore2 from time to time to generate XMLs and it is doing a good job if the source is (print and scan) is high quality and the music is not too complex. For example, if there are multiple voices in one staff things sometimes get messed up. In my opinion, best results are achieved, when you use a PDF/JPG as source which was built by a notation software. I had to convert some (easy) piano accompaniment for a solo piece some weeks ago which I only had PDFs of. The results were absolutely usable without lot of manual editing in Dorico.
I ran a couple of English song cycles (Stanford and Finzi - clean PDFs) through PlayScore 2 earlier this month, specifically for the piano parts. It’s not bad at all but didn’t necessarily save me much time: I’m quick at note entry.
In my (albeit limited) experience it gets note values and pitches mostly right, picks up articulations and dynamics mostly correctly, but is stymied by grace notes and sometimes gets confused by where slurs start and finish. The most frustrating situations are where PlayScore gets a rhythm wrong: it’s quite possible for the remaining bars of the score to have every note a hemidemisemiquaver (64th?) out from where it ought to be. Tuplets are sometimes the cause of this. Sometimes these shifts affect different voices in different ways, which I guess isn’t too bad if each stave is homophobic but is a pain if you’ve got five or six voices across a piano grand staff.
Given PlayScore doesn’t have any editing features at all, all the clean-up has to be done within your chosen notation software. Dorico has Insert mode; Sibelius doesn’t (and its sliding/shifting functionality is rubbish compared to Dorico’s).
Well, Andreas, it’s exactly the same with Photoscore. It reads almost perfectly any pdf score output by a recent program (Lilypond, Sibelius, Finale or Dorico)
Problems start with dense scores (with condensed instruments, triplets hidden, multiple voices, cross-staff voices…)
There’s still quite some room for improvement with those programs.
Ever since music recognition started I’ve supported the effort by buying or upgrading the software and trying it out, and each time I come to the conclusion that I could have entered everything from scratch more quickly and more securely, but then, like Leo, I’m quick at note entry.
Sorry you have not had good luck with OCR, Vaughan. I use SmartScore Pro exclusively for note entry when I am editing standard literature. It is so much faster than I could possible enter it by any other means.
Hi, this is PlayScore here. We saw the messages about our app. We would be pleased to offer a free subscription to someone interested in reviewing PlayScore 2 for iOS on forum. You can contact us at firstname.lastname@example.org.
The iPhone version of PlayScore get a 4.1/5 but it’s based on a mere 44 reviews. The one-star reviews may well stem from problems with the images that users have submitted to the software rather than the software itself.
I just tried the free version of PlayScore Pro for Android. I don’t see that they have released PlayScore 2 for Android yet.
The good news is that the app doesn’t crash and is pretty obvious to use. But as for as scanning, it is beyond worthless. I tried 4 different pages. All were single instrument (solo) with single melodic line. Two were hand-written. Two were nicely engraved. None of them did better than about 5%. The last one was the cleanest, simplest thing I could find. PlayScore could only recognize about 1/10 of the measures. There were 3 measures. It plays pack 4 random measures and stopped – and it didn’t play them anywhere close to correctly.
It is a fantastic concept – literally. It is pure fantasy at this stage. Maybe with incremental improvements, it might become useful. I hope so. It is a dream, just not a dream come true yet.
I had this argument with the CEO of Musitek. “The software is fine, it’s the images that are wrong.” The lines weren’t strong enough and breaking up. Not enough contrast. Too much contrast. Too much dust, hairs, etc. The image wasn’t square. The image had no grey tone. Yada yada yada.
One of the main reasons for scanning something is because we want a nice clean engraved page instead of the crappy old sheet with broken lines and odd notation conventions that’s been photocopied 40 times.
Don’t show me what you can do with perfect notation on a pristine page: I’ll stump up the $400 when you can do a decent rendering of a crappy image. And when you’ve worked out that 12 equal notes in 4/4 probably means triplets.
The tech industry invariably proceeds through a succession of annual buzzwords. This year, one of the prime buzzwords is “artificial intelligence”. Although I don’t see the PlayScore people hammering that term, I do see lots of software vendors doing that, and it becomes marketing nonsense.
The basic standard for computer “intelligence” is, if a human can look at it and immediately makes sense out of it, then a computer that has “intelligence” should be able to do the same thing. But none of the platforms people use for this AI stuff has anything at all that integrates knowledge into the process. In your example, if there are 12 eighth notes in a 4/4 bar, then we can be pretty certain it is 4 sets of triplets, even if we can’t clearly see the triplet designations. None of the AI stuff works this way. It is just pattern marching, not “intelligence” per se.
In all those 4 examples I just tested, I can look only at the photo (not the original paper) and instantly understand every note. There is no ambiguity. Yet it vexes all these scanning programs to one degree or another – to a very high degree in the case of PlayScore.
As much as we all wish that computers could “think”, “reason”, and “apply past experience to the present situation”, they can’t. This is why only a fool or person with a death wish would trust a car to drive autonomously in heavy traffic. The technology isn’t nearly as far along as we wish it were.
I hope to live long enough to see a car that can safely drive me anywhere. And I hope to live long enough to use a music scanning program that works. Meanwhile, let’s all appreciate that Dorico isn’t artificially intelligent, per se, but can save us all enormous amounts of time while producing output that makes us all look good.