Did you read the Please Read and Guidelines on the render page? For example do you have divisi? Check hyphenation. Are you able to simplify it somehow?
Otherwise try again tomorrow or so. It seems to be constantly updated as they fix bugs and add improvements; the server is rebooted from time to time because of various problems …
Thanks @arco, the issue wasn’t with cantai, it was that the musicxml file exported from Dorico seems to have only exported fragments of the lyrics. So when I imported this file into MuseScore the lyrics were incomplete. It only affected the soprano part though as when I later exported the alto, tenor and bass parts the lyrics imported into musescore fine. Very strange. I can’t be certain as to whether it was an export bug from Dorico or an import bug from musescore for that part. I was just curious as to whether anyone else had similar xml export/import issues is all. FYI I am using MuseScore to generate the midi file that can then be rendered in cantai.
That’s odd. Did you try deleting all the other instruments/voices leaving the soprano, then export xml, just to see what happens? Does it make any difference when imported into MuseScore?
Could you easily delete the lyrics and re-input them and see what happens? Also perhaps making sure they are in the same voice as a possibility?
I have been doing the same as you. I sometimes have had to tidy up in Musescore, before Cantai renders but often no problems, but have not done much, I’m really waiting for more bugs to be fixed like, as you know, pronunciation.
No problems using UtaFormatix.
Here’s my route again, if someone doesn’t know this already: Export MusicXML from Dorico and import it to UtaFormatix, export Standard MIDI [MIDI – Mid (Standard)] from UtaFormatix and upload the MIDI to Cantai Web Renderer. No MuseScore in my case, that is. I find this way just a little bit more comfortable. 
UtaFormatix is a fast and convenient web application for converting projects among singing voice synthesizer softwares such as VOCALOID, UTAU, etc. So no need to install (and learn the [easy] use of), say, MuseScore, if you haven’t got it already.
All of the above, because Dorico does not export lyrics with MIDI, but of course does with MusicXML.
Someone (David Das) posted this demonstration process from Dorico through UtaFormatix to Cantai.
Meet Cantai: The NotePerformer-style revolution for AI-powered choral music
I see they are now welcoming educational price enquiries:
They are working on musicXML import into the Renderer.
I’m posting this here since the workflow will be pretty near identical in Dorico.
Watch this step-by-step demonstration of the Cantai workflow in MuseScore. The video shows how to install the VST plugin, assign Cantai voices to vocal tracks, render the score, and make changes to the music in real-time. Future versions will automatically watch for changes and rerender as needed, with improved file management for easy export to other audio applications.
Absolutely wonderful. Thank you so much, Richard @turingopera.
That’s a good demo, thanks!
Unfortunately, I still haven’t managed to get Cantai to render any files and I can’t find the solution.
You should get in contact with the people on Discord:
That’s a pity – and strange. While waiting for the VST, I’m rendering all the time the way I’m describing here:
this is interesting but I’m not sure what it means in practise. Suppose I have a 30 minute choral piece – will the renderer automatically update as I go along in my so there is not a long wait every time I want to hear what I’ve just composed? If it’s possible to hear in real time - just as with instrumental work, what’s just been written, that would indeed be a considerable step forward and indeed better than for instance EWQL Wordbuilder where you need to put the text at the top or start from the beginning each time (strictly speaking,I believe there are convoluted ways round this but nothing I’ve found practical)
Right now it seems like people are more interested in ethical and philosophical discussions than in solving the question of how to make a working midi (that’s what I read on Discord)
By the way, I’ve tried step by step the methods you’re recommending and I’ve even copied the scores from the program’s creator and I don’t know what the problem is, but it doesn’t work for me. Neither Doric nor Musescore.
Happily, that’s not true at all. They are working very, very hard behind the scenes to make not only MIDI, but especially MusicXML work (directly from Dorico and other scoring programs to Web Renderer). And the VST alpha for MuseScore is almost here, test group getting it hopefully already this, or maybe next, week. VST then later for Dorico and Sibelius.
Be patient, please, everything is progressing just fine.
Really a pity. Works technically ok with my MIDI files (originating from Dorico) for me.
Interesting developments. Out of curiosity: I see that (American) English is becoming available, and that Latin (but in whose pronunciation?) and Spanish (also quite some variety) are in preparation. That’s huge, but it still doesn’t cater to the several thousands of other languages that can be found on this big planet of ours, let alone their countless accents and dialects. I guess it’s unlikely Cantai will develop models for each possible language, but what if it would support IPA (International Phonetic Alphabet), the linguists’ choice for describing human speech in detail? You might be able to fairly reliably construct any sound like it occurs ‘in the wild’, provided you have knowledge of IPA, of course. In that case, one might write any language in any peculiar accent, without having to wait until Cantai supports that particular variety.
Perhaps post a (cut down) very short version of the Dorico file and maybe someone here will see if they can get it working as an initial test?
When I had problems, I would export/render each vocal staff or portions of it individually so I could locate what might be wrong and where. Sometimes it was hyphenation. My tests have been quite basic so far because it’s still in beta.
Where I had used “mm” as a word an error message showed. I think I read these (mm, ah, oo etc.) will render when xml works.
These are the Guidelines on the render page, presumably you have gone through these?
Guidelines
But above that, in the Please Read section is their “Contact us for support” link. If you have problems, they would presumably want it to analyze so it can be fixed if it is a problem at their end.
Email me the details: cantai@turingoperaworkshop.com or join the Discord Turing Opera Workshop