export a proper formatted score from Dorico as mxml to see what this file has in place of these parts. Try transplanting that part from the proper file to the “bad” one and see if this helps.
Ok I’ve removed the offending bars from the Photoscore project as suggested by @_TRG and lo and behold the import is absolutely fine in Dorico/Musescore
Conclusion: Don’t believe Photoscore when it tells you that the timing in bars (especially with multiple voices) is correct!
So if someone sees this thread in future and has trouble going from Photoscore to Dorico: a possible solution is to open it in Musescore first to see if there are timing errors in the XML file. Hopefully Dorico will be able to list these errors itself at some point in the future, which would save this intermediary step.
Thanks to everyone on the thread for their suggestions and help, this community is awesome
Usually this only happens if photoscore has mis-read something and you have corrected it. eg. Photoscore decided to add extra voices, or put part of a voice 2 tuplet into voice 1. I’ve only ever had this problem in bars with multiple voices and tuplets.
I’ve never seen photoscore change the page dimensions before - what were the dimensions of the original pdf, or was it a scan?
It shouldn’t make mush difference, but photoscore only produces xml v2.0 scores. I think Dorico works on v3.0 and the latest ‘standard’ is v4. Hopefully photoscore will catch up.
Hmmmm. There’s something in the back of my mind that tells me there is some relation between the dpi (resolution) of the scan and assumed staff height, etc. I only very vaguely remember this.
Don’t even know if it was a Photoscore thing or some other software.
But I think there’s some relation, and “wrong” dpis can throw the OMR off, and probably mess up the layout?
Maybe check the documentation for such issues and/or recommendations, and check resolution of the scanned files ? (Or whatever image it is that is fed into Photoscore)
Photoscore now provides 3 different recognition algorithms supposedly for low, medium and high definition scans. I’ve had mixed results with these as success also appears to depend on the physical size of the original and the amount of noise (spurious dots and marks) on the page.
Usually I will test a typical page using all three, to determine which is likely to give the best (ie fewest corrections) result for any particular score. That said, IMO, photoscore is still the best OMR tool available, despite its quirks and limitations.