Transcription support using moveable window on source image
I will describe a fairly simple mechanism that I suggest to support transcription from an image (pdf, jpg, etc.) of sheet music into MuseScore. Many people (such as myself), often transcribe from an image on the computer. That image may be from a scan of sheet music done by oneself, be purchased, come from IMSLP, or from wherever.
Usually, it is of paramount importance that the transcription is accurate. And it is easy to get confused and make mistakes during the transcription.
The idea is that the user has a resizeable thin-border rectangle containing part of the original image, and she moves it around to line up right above or below the transcribed measures in the target MuseScore score. Suppose you try making an Adobe Acrobat window a couple of inches tall and about 6 inches wide, and placing it right under the relevant measures in MuseScore. The obvious problems are: 1. The borders of an Acrobat window are clumsy and wide; 2. The moment you click on MuseScore, the pdf display window disappears under MuseScore. At times I've had to resort to holding paper scores up against the monitor to check a complex section.
Here's what I suggest, described in terms of the user experience:
1. The user opens her target .mszc score and selects "transcribe from image" and selects a source image/pdf/etc.
2. Assuming a typical layout on a 67cm (27") monitor, She sees a default 5cm x 15cm (2" x 6") window onto the start of the pdf, with a 2 pixel black border. The source image window is placed by default in the middle of the open .mszc document.
3. The user resizes the image rectangle via a corner in standard fashion. She moves the rectangle containing the origin pdf to just under or over the target bars of the MS score, lining it up. She scrolls the image, and/or holds the mouse down and drags the image around inside the fixed rectangular border, so that the section of interest is visible, like you drag Google maps around. She zooms to make the current source image bar the same size as the target bar.
4. The user highlights the target bar(s) of interest in the MuseScore score, and likewise highlights the corresponding section in the original image score. Highlighting can be done by a shading the bars in some color, or enhancing the barlines either side of the bar of interest with a thick yellow line.
5. When the user clicks and edits in the MS score, the inset original image remains visible (does not disappear under!). This part is crucial.
6. When a bar is finished, the user can mark the target MS bar and the original pdf bar as finished. Possibly using a colored horizontal line, or a tick/check.
7. When the user ends the session, her places in the MS score and the pdf are saved, as are the marked off finished bars. Of course, this can get messed up if the user edits the file later without using the origin image.
Final remarks. This enables the original and the target to be as physically close as possible. This is a first shot at describing it, and I expect that if it were done, there would be many changes. It all seems rather childish perhaps, but let's face it, it's a real problem, and lining things up is really useful. Software developers do it all the time, for example in diff tools. Perhaps MuseScore could "score" a first in this regard, as I don't think any other notation programs do it. What do you think?
Comments
Some work on a similar feature has been done during the development of MuseScore 2.0. I believe it covers 1, 2, and 3 except it does it automatically by detecting staff lines and barlines. Unfortunately it's not ready for prime time. You can see a demo here http://www.screenr.com/g1Bs
Your last remark is not going to happen anytime soon. MuseScore is not a drawing software where we could draw exactly the same than an original score. It's a score writing software which try to encode the best practices in notation and so end result will always be different to your original if you compare them pixel by pixel. That's also why it seems better to have the PDF/image on the side instead of below the score view.
Thanks Lasconic, I was not aware of that nascent OMR feature. Sorry if I should have found it. Yes, it would help a bit in some pdf cases, and I would definitely use it. I watched the Screenr. My idea of course had no OMR in it.
I think there is some confusion. I was not talking about reproducing the look of an original. I'm just talking about normal transcription. For example, recently I transcribed a violin duet by Vaclav Pichl from a book on imslp.org. The original was in hand-written notation (or it looked like it) and of course the MS version was very different. Most relevant here is that the original had the Primo and Secondo in two separate parts, very densely packed, hard to read, and with different numbers of bars per line. There were inconsistencies that had to be resolved. I put these into MS in score format with a pair of staves. So any normal OMR would have been flummoxed. It was quite confusing, and I had to hold a printed Pichl original against the screen often. To check the final version I had no option but to print both the originals and the MS version out, write some bar numbers on the originals, fold the sheets as needed, and put one on top of the other, so that I could get the matching bars physically as close as possible. I can upload my score, etc., but you get the idea. It's a primitive solution in a sense, and primitiveness correlates with robustness and applicability.
I realize that there are many features in the backlog, and that this will probably not get done, but I'm just floating the idea. Trying to be a good citizen and all that. I'm sure that a lot of people would benefit from it, including potential professional editors. I don't hear any support though!