The new MusicXML import feature in 2.2.0 is fantastic — having good importing of multi-syllable words prompted me to buy Synth V.
I made a folder with MusicXML and MIDI files (exported from Sibelius) and .svp files imported straight from those, which illustrate two issues I see in 2.2.0b2: [link to Synth V test files folder]
A Tied-Note Reading Test
One mistake I see is this subtle one: Notes that are *tied* together should be treated as one syllable, but in MusicXML import it looks like they are being treated as two notes of the same pitch, slurred. Slurring from one note to the same pitch is rarely seen in sung music, but it’s actually quite nice how Synth V handles the subtle effect. Import from MIDI handles the tied notes perfectly (treating them as long notes), but from MusicXML it treats them as slurs. Here’s the lyrics I used:
These slurs sound good
These long notes, too
These tied notes should sound like long notes
These slurred notes sound good but those last tied notes should not sound like these
A Trickier Reading Test
On the other hand, multi-syllable words are handled well by the MusicXML import but not by MIDI import (this has been a problem since v.2.0). Check the pronunciation of the word “double” which comes out as if it were two words, “dou” (doo) and “ble” (bll).
Thanks for the great new features, and I hope these two little quirks can be addressed.
I have found the same kind of things. It all depends on how the Music XML file is notating slurs and melisma. Not all files are consistent, and Synthesizer V needs absolute consistency. I found that some long notes were interpreted by Synthesizer V as several notes with a dash “-” between them and if you look at the wav-form you will see that this affects the sustain of the note with an unpleasing change of pitch. So the simple solution I found is to solo and play each track in Synth V and watch whenever it does that and manually “merge” the two notes (which I have assigned to a single key.)
This is not really a Synthesizer V problem which they are able to fix. It is an issue for the makers of the XML, which needs to be standardized in a way that works with Synthesizer V. I will learn how to minimize this in making my own XML files, but a little listening and tweaking in Synth V is still a lot less work than before, so I can’t complain.
Likewise, you need to listen to each track in Synth V to get the pronunciation right on every word, because there are sometimes inconsistencies in third-party downloaded XML files. Also, XML files often have punctuation like, commas, periods and exclamation marks after word, which makes them unrecognizable to the dictionary, so if you have any of those words in your dictionary, Synth V will pronounce them wrong. I wrote to Dreamtonics and suggested that they eliminate punctuation from the XML import, because Synth V does not need them to produce voice tracks. So in the meantime we just need to listen and correct these whenever needed.
The point is, that with a small amount of tweaking you can get great results.
Hi Thomas,
Thanks for your input, and I know that this can be done right. I’ve been using the very consistent Sibelius-exported MIDI and XML files with another singing synthesis system for some years now, and correctly handling ties and slurs have not been a problem with that software (link, if you’re curious).
After all, in Synth V 2.2.0b2, the MIDI import works almost perfectly and the MusicXML import works almost perfectly… I’ll give the developers the credit to be able to address both of these issues.
–Dick
For the tied notes issue I hacked quickly together a small script SynthV2-Scripts/MergeTiedNotes.js at main · egonelbre/SynthV2-Scripts · GitHub . If you encounter any bugs, let me know… although I hope this will be fixed for the release.
Another problem - When I transfer data from Dorico to Synth V using MusicXML, the apostrophes get converted into some sort of curly apostrophe in Synth V. I have to manually edit all of them to avoid strange pronounciations, gaps and hiccups.
–Neil