stakira
Add MusicXML Import
Adding MusicXML capabilities may be the starting point for future support with NEUTRINO. This PR is for importing.
XmlSchemaClassGenerator used to generate MusicXML 4.0 schema for use with XmlSerializer. Most of new lines of code comes from it in: MusicXMLSchema.cs
Corresponding Discord Thread
Pitch, Tone, and Duration fields being read.
Thinking now about how key signatures, time signatures, and whole/half/eighth/sixteenth/etc. notes interpreted.
Each individual notes includes information under the <alter> tag when exported:
https://www.w3.org/2021/06/musicxml40/musicxml-reference/elements/alter/
Verified under MuseScore by examining the musicxml export under different key signatures. As such, keeping track of key signatures is unneeded.
Also for reference: https://www.musicxml.com/publications/makemusic-recordare/notation-and-analysis/a-sample-musicxml-encoding/
There exists compressed musicxml following the zip compression method under the file format .mxl but that can be addressed as separate concern.
Addressing oxygendioxide's concern on character encoding, remember to check for it. Can see for an example: https://github.com/stakira/OpenUtau/blob/b969d52860bf03502418b0f9353723c9f61d0450/OpenUtau.Core/Format/Midi.cs#L32
Check the entire .musicxml file to determine character encoding. It is assumed if the lyrics have a specific character encoding, so shall the overall .musicxml file and vice-versa.
Regarding length of time, the <divisions> and <duration> are linked to one another: https://www.w3.org/2021/06/musicxml40/musicxml-reference/elements/divisions/
Seems to be a viable way of interpreting note length. A paper published on 2021 provides details on the <duration> element and a survey on how it is used in various programs: https://musescore.org/sites/musescore.org/files/2021-05/TN-MXML-duration-overview-i01.pdf
The <type> element may be an alternative usable with the single dot <dot> element and double dot <double-tongue> element. https://www.w3.org/2021/06/musicxml40/musicxml-reference/elements/type/
Went with <divisions> and <duration> route for interpreting note length
Importing a basic MusicXML with Open... or Import Tracks... completed.
Using a sample MusicXML provided by NEUTRINO, importing into OpenUTAU, and rendering with Teto appears to match the result of the render from NEUTRINO with Zundamon
Valid skeleton MusicXML being exported. This is the export being opened in MuseScore
Now to figure out exporting notes.
Export compatibility is in context of NEUTRINO. MusicXML is actually only used as an intermediary format to be converted to HTS .lab with a bundled utility named musicXMLtoLabel according to the the script which runs NETURINO:
The utility appears to be derived from https://github.com/taroushirani/xml2lab which itself seems to draw from https://www.sinsy.jp/
Food for thought, perhaps it's possible to skip straight to the .lab format.
Able to have no elements under for full rest like <measure number="1"/>. Measures don't appear skip-able in number however
Order that elements occur in a measure seem to matter to NEUTRINO
One would think generating MusicXMLSchema.cs with -r option to define order XmlSchemaClassGenerator would work. Turns out it's way too strict requiring deserialization to be in the same order. Even though order matters, MusicXML does not require strict ordering so this cannot be a solution.
One way to modify order on serialization is rearranging the order of the definitions themselves in whatever class within MusicXMLSchema.cs. Kinda hacky although
There might be some incompatibility between programs. OpenUTAU and pals allows time signature change basically anywhere. MuseScore only allows placement at the start of the measure and following from that I dunno whether this is acceptable in the MusicXML standard.
Related to previous comment about order of elements mattering(?)
For sake of simplicity, leaving out support for tempo and time signature changes in the middle of a measure.
Future changes add it can probably rewrite/customization serialization behaviour to do stuff like target NEUTRINO
Use GetMinDurTick or related to get minimum length. Can use that with BPM and time signature number to calculate the total number of measures to export.
Use dynamic programming to efficiently translate OpenUtau notes to optimal number of MusicXML notes. See coin change and/or knapsack problem for reference. Data structures and algorithms coming in handy out in the wild kinda surprising :thinking:
Splitting Import/Export into different PR's to reduce complexity and have Import for testing earlier. Export will be handled in another PR
I added some unit tests for this musicxml importer.
https://github.com/oxygen-dioxide/OpenUtau/commit/23b9183adefebb62508512cf478daf7819f00a0b
Musicxml files come from cuthbertLab/musicxmlTestSuite
Thank you!
On Mon, Dec 16, 2024 at 20:49 oxygen-dioxide @.***> wrote:
I added some unit tests for this musicxml importer.
@.*** https://github.com/oxygen-dioxide/OpenUtau/commit/23b9183adefebb62508512cf478daf7819f00a0b
Musicxml files come from cuthbertLab/musicxmlTestSuite https://github.com/cuthbertLab/musicxmlTestSuite
— Reply to this email directly, view it on GitHub https://github.com/stakira/OpenUtau/pull/1152#issuecomment-2547390019, or unsubscribe https://github.com/notifications/unsubscribe-auth/BFDBW7U5EV6NRGAYQXBAD4T2F6GM3AVCNFSM6AAAAABI73ZCKOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNBXGM4TAMBRHE . You are receiving this because you authored the thread.Message ID: @.***>
@oxygen-dioxide Rebased onto the latest commit from master and then applied the commit from your branch on top.
On Wed, Dec 18, 2024 at 13:42 四十二porime @.***> wrote:
Thank you!
On Mon, Dec 16, 2024 at 20:49 oxygen-dioxide @.***> wrote:
I added some unit tests for this musicxml importer.
@.*** https://github.com/oxygen-dioxide/OpenUtau/commit/23b9183adefebb62508512cf478daf7819f00a0b
Musicxml files come from cuthbertLab/musicxmlTestSuite https://github.com/cuthbertLab/musicxmlTestSuite
— Reply to this email directly, view it on GitHub https://github.com/stakira/OpenUtau/pull/1152#issuecomment-2547390019, or unsubscribe https://github.com/notifications/unsubscribe-auth/BFDBW7U5EV6NRGAYQXBAD4T2F6GM3AVCNFSM6AAAAABI73ZCKOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNBXGM4TAMBRHE . You are receiving this because you authored the thread.Message ID: @.***>