ScanScore seems to be the only music scanning program that us under significant development. I wonder if anybody has used it enough to form some opinions, in particular how it compares with the older programs (SmortScore etc.)
I have been using SmartScore and also PDFtoXML (which isn’t really a scanning program). The results are only so-so, and scanning scores of transposing instruments seems somewhere between tedious and impossible.
I just wonder if ScanScore has any better prospects.
Photoscore Ultimate is supposed to be upgraded in November… I would not say it’s really under development, but it’s not totally dead 
I was unimpressed by ScanScore’s trial, so I didn’t buy it. There’s no list of what notations it can recognize, but it seems to miss quite a few. But you can be the judge of that – the trial is fully functional, and only prevents you from exporting.
PhotoScore has been pretty good. The interface takes a bit of getting used to, but otherwise it’s pretty quick to make timing corrections. The rest can easily be fixed in Dorico.
There’s also the open-source program Audiveris, which is under development. The latest version employs machine learning, and it is possible to train it with your own symbols to improve recognition. Kind of hard to use right now, but it’s really exciting stuff!
Thanks for the information about Audiveris, Jester! I’ll check that.
Make what you will of the fact that the Photoscore website announced the Catalina update at least 16 months before it was supposed to happen. (the “date last modified” on https://www.neuratron.com/ says July 2018 - not July 2019!!)
The “About Neuraton” page hasn’t been updated since 2015.
The “Product support” page hasn’t been updated since 2011, and still has a comment about support for Sibelius version 3.
“It’s not dead, it’s resting” ??
Dear Rob,
The last update from neuratron was really on July 5th, 2018. I don’t understand your sentence. And we’re expecting an update (upgrade?) but I confess I do not have high expectations on this (apart from Catalina support…)
The “Date last modified” line at the bottom of a web page usually means when the page was edited (for example so that search engines can discover when the page was updated).
If seems remarkable that they would announce an update for Catalina, more than a year before Catalina was released.
Of course the “July 2018” date might be a typo for “July 2019” but a modern website would insert the date automatically.
It just gives me the impression that the website isn’t being actively maintained by anyone.
Ok, I understand. For what it’s worth, the announcement for Catalina has been added mid-october. And you are right, the date must not be automatic on this site, and it indeed looks quite old and not maintained…
In my experience, the biggest problems show up when importing. I can have the music looking good and even playing correctly under the scanning program. But when importing into the notation program, it can be a complete mess. And it can be a DIFFERENT kind of mess in each notation program. And then if you can clean that up, there are often lingering problems, to the extent that whenever somebody has some strange problems with a score here, it is common to ask them “Did you import this from XML?”
A free trial without an export capability doesn’t do me much good. These publishers should find a different way to restrict their trial versions.
cparmerlee, you didn’t say, but perhaps you use SmartScore. All I expect from SmartScore X Pro is note input and would never do any substantial editing within SmartScore for the reasons you mentioned. But the program has saved me many hours of needless work and for that alone I think it is worth every penny I paid for it.
When it works, it is worth every penny. When it doesn’t work, it takes years off my life. But my main point was that I can live with the accuracy of scanning it delivers. Most of my meltdowns are in the integration (i.e. import) into the notation programs. And again, it fails different ways in different programs. SmartScore doesn’t usually create really wacky notation when imported. PDF2XML can create absolute messes. I think my biggest frustration with SmartScore is that I really can’t use it on scores that include transposed instruments because it simply doesn’t recognize that a BB trumpet is not a concert instrument.
I find my only real option is to SmartScore one instrument at a time if I have parts available. But then measures get all mis-aligned because of multi-measure rests that are read wrong or bar lines that are missing.
And if I get past all of that, there invariably is rubbish in the Dorico (or Finale ) file that causes other notation problems. For example, on my latest project, the whole thing is 4/4 time, the meter is getting lost somehow. If I don’t force in a 4/4 on every measure, then you get the bad notation that is seen in in measure 44, with a quarter crossing the half-measure. I have tried every layout option and the only way I could get that to work is to force in a 4/4 every affected measure. For some reason, Dorico doesn’t believe the time remains 4/4, but there are no signposts that would indicate there are any hidden meter changes. This is the kind of thing that makes the XML input so tedious.
Another example on this same file is that all the measures are shifted off by two, as if there are two phantom measures at the beginning. I added a bar number change on the first measure, and that seem to have no ill effects, but is just one of those rough edges.
I eventually gave up and created a clean flow, cutting and pasting everything from the damaged flow. Fortunately whatever garbage was deposited by the XML was not carried across in my cut-paste operation, so I ended up with a “normal” Dorico flow. But it feels like navigating a minefield.
While I am not scanning orchestral scores, I am able to OCR complex piano music in SmartScore without such errors. For example, it recognized Beethoven’s 32 piano sonatas with only a few hitches. It does leave out occasional bar lines, but these are easily inserted within SmartScore.
So this makes me wonder if you using the stand-alone SmartScore X Pro or later. I found lite version in Finale impossible for my purposes. And I wonder also about the clarity of what you are scanning.
I am using the pro version X2 and the latest patches (which are few and far between). I am not really unhappy with the scanning part. it works well enough to be useful and I am reasonably adept at making minor corrections within SmartScore. The big problems happen when importing. I don’t doubt that it would do OK with piano scores. When you toss in transposed instruments all %$% breaks loose.
And to be honest, I’m not positive whether it was SmartScore or PDF2XML that caused the meter-related problems I mentioned above. I was frantically trying different combinations of both programs, and they each had their own set of obstacles. I am going to try two more very similar scores this weekend and will try to take better notes.
I just OCRed the first page of the Rachmaninoff Symphony no. 2 with SmartScore and imported it into Finale 25.5. I was impressed with the results. I didn’t have to do any editing in SmartScore and only a little in Finale. One would have to redo everything besides the notes, however, but I expect that.
I also tried the same MusXML file in Dorico. For some reason, all the bar lines were misplaced, throwing everything off. I found this surprising because piano music has imported so well into Dorico.
I then tried another more complex page of the score. This produced more rhythmic errors in SmartScore, which I didn’t correct. It imported into Finale less well, but remained reasonably intact. It went bonkers in Dorico.
From this very cursory view, I wonder if MusXML files must be rhythmically correct for Dorico to deal well with them. Finale does not seem so thrown off by a few rhythmic errors. Perhaps this is because Finale is so measure-oriented, and Dorico is not.
The fact that for decades, Finale itself couldn’t automatically enforce having the correct number of beats in any bar might well be an advantage reading files where the rhythm isn’t quite right
MakeMusic also has had the inventor of MusicXML on their payroll for a long time, and presumably that is a measure of the amount of software development resource they have dedicated to supporting MusicXML without any distractions from other priorities.
I did another project with SmartScore this evening. This was a fairly simple saxophone quartet. These programs can’t handle transposed instruments in a transposed score. So I had to do it one instrument at a time, and then transpose that material to get it into the right key once it is pasted onto a Dorico score properly built for transposing instruments. That’s fairly laborious, and negates most of the benefit of scanning.
Once the material was into Dorico, there were still lots of problems. The pick-up note was on beat 1 of the pick-up measure, not best 4 where it should have been. Lots of measures failed to display rests (implied or otherwise,) so I had to fiddle with that – basically I had to add notes and delete them to get Dorico to start displaying whole rests.
One of the more annoying things is that cautionary accidental rules don’t work at all. It seems that the cautionary accidentals came through SmartScore and Dorico treated them as if regular accidentals, and therefore wouldn’t obey the parenthesis rules. I had to change all of those individually in properties.
The point here is that the scanning is acceptable in many cases, but the process falls apart in the hand-off through XML.
I have had great succes with NotateMe on iPad (using the photoscore engine).
Mostly SATB scores, for the purpose of making audio rehearsal tracks of existing arrangements.
Sure, it was last updated about 2 years ago, but it saves a ton of effort having to type everything by hand (I started using it when it was still in beta so I got it at a discount).
It exports MusicXML which I then ‘correct’ in a notation program (depending on the quality of the score I hardly need to do that)
One added advantage for me is that I can do this also while traveling with only an iPad (e.g. when getting new sheeetmusic in a workshop weekend).
Craig, do pay attention to your MusicXML import settings within Dorico. There’s a possibility you could have saved yourself significant time by unticking the “accidental visibility” tickbox. As for rests not showing, that’s generally down to automatically set Starts/Ends Voice properties. There’s a way or selecting all and setting the starts/ends voice (to off!) en masse, too - I can’t remember what you need to filter off the top of my head.
cparmerlee. I am afraid that my initial good impression of Dorico input direct from SmartScore was an error, since I was actually importing Finale MusXML files into Dorico.
I just input a solo piano movement into both Dorico and Finale, each directly from the same SmartScore MusXML. The Dorico was a mess and unusable. I did have all options checked in the Dorico MusXML preferences. The Finale just needed the usual small corrections. Then I converted that uncorrected Finale file into MusXML and imported into Dorico. The result was quite like the Finale and very usable.
So that would be the path I would take were I using Dorico as my main notation tool and SmartScore as my OCR tool: SmartScore>Finale>Dorico.
It seems to me it would be in the interests of all the scanning vendors and all the notation vendors to get this straightened out. Everybody’s product is more valuable if it can be more or less seamless in this regard. I don’t want to minimize the complications inherent with MusicXML, but it does seem that products like Finale and Reaper are doing substantially better than Dorico.
I repeat myself: I can live with the state of recognition. Of course I’d welcome any recognition improvements, but the big time-wasters are in the integration side.
