Klarinet Archive - Posting 000075.txt from 2002/10

From: LeliaLoban@-----.com
Subj: [kl] OCR software for scanning music
Date: Wed, 2 Oct 2002 09:15:34 -0400

Robert Moody wrote,
>I suggest that Finale does the job, but Sibelius 2
>with PhotoScore does a much better job. And
>though not 100% still, it DOES interpret the
>dynamics and other articulation markins as well
>as instrument labeling so that the scanned products
>acts and responds like you typed it yourself.

PhotoScore works that cleanly for some scores, but not all. It's fine for
converting, say, a Bb clarinet part to an Eb clarinet part. The program
handles a lot of obvious potential problems. It can scan from a bound music
book that must be laid out on the scanner table with half the pages upside
down, then flip the scans right-side up on command, for instance. However,
computer scanning of music is still new technology and work in progress. The
more staves to the music and the more voices per staff, the messier things
get. Caveat: I'm using the PhotoScore Lite version included with Sibelius
2.1. Maybe PhotoScore Professional (available as an upgrade for more money)
handles complex scores better, but I'll wait for a few more layers of
upgrades and reports from happy users first....

Yesterday afternoon, when I decided I might as well try my own suggestion
about transcribing Couperin's "Les Baricades Misterieuses" for four clarinets
as an opportunity to learn more about working in Sibelius, I was unable to
make a usable (in the sense of practical) scan with PhotoScore Lite. This
keyboard score on two staves is heavily syncopated, with a lot of notes tied
over the barlines and a lot of note stems shared by more than one voice. The
computer program can't always sort out what's going on. Although the program
reads the pitches of all the note heads correctly, *every* measure in the
entire piece scans in with an incorrect number of beats (some plus and some
minus)!

PhotoScore knows when it screws up. The program left me a neat red warning
mark at the end of every bar to indicate that it needs fixing. Yowie, that's
a lot of red. Worse, from a human point of view, the program's errors often
look counter-intuitive. The prospect of correcting every measure and *then*
converting the score from keyboard with two bass clef staves to clarinet
quartet on four separate treble clef staves doesn't thrill me. After
selecting a brief passage and experimentally correcting and transcribing it,
I said, "The hell with this." Extracting the two voices from each keyboard
staff in Sibelius *puts back* some of PhotoScore's errors, while adding a
fresh new layer of errors!

Probably much of the fault lies in my inexperience with PhotoScore, but in
this case, working from a keyboard score of only three pages, I'm finding it
faster and less frustrating to read the printed score with human eyeballs
while inputting the notes straight to the clarinet staffs from the computer
keyboard from scratch (and make my own mistakes...), instead of figuring out
all those computer errors one by one. Maybe I just need more time and
patience with PhotoScore--but not with a score that the program messes up
this badly. I'd also like to get good enough at playing metronomically to
input into Sibelius with MIDI from the electronic piano keyboard (one voice
at a time, each into its own clarinet staff). So far I end up with a lot of
rhythm errors that way, too.

Lelia

---------------------------------------------------------------------