Open source singing synth.

[onefang]

For various reasons I'm wanting to use a speech synth to do my singing for me. One of those reasons is that the last time I did any performance singing was several decades ago, my voice is very rusty. Everyone else's second hand tobacco smoke seeping into my room isn't helping my asthma nor my singing.

Most of what I have found seems to be based on closed source Windows backends that do the actual singing, used via WINE. Even though there is open source front ends, I don't want that, I want fully open source and Linux.

Festival has a singing mode, but it only works with the low quality voices. I want something better quality. The festival docs keep mentioning flinger for better quality, but that's no longer part of festival.

Are there any other fully open source alternatives for Linux that don't need WINE? For bonus points an Aussie accent voice.

[khz]

Spontaneously Xmonk https://github.com/brummer10/Xmonk.lv2 and XPolyMonk https://github.com/brummer10/XPolyMonk.lv2 from @tramp come to mind. When you play the synthesizer with the Pitch/Mod wheel it starts talking.

Possibly also feasible with Cardinal https://github.com/DISTRHO/Cardinal.

[onefang]

A quick look at Xmonk and XPolyMonk and nothing screams "insert lyrics here", the only voice thing I can see is "MODWHEEL CC 0x01 -> vowel parameter". Neither is available in any of the Devuan repos I'm currently feeding from. So I'd have to compile them. I'm OK with that, I am a very experienced computer programmer anyway. So long as I don't have to jump through too many hoops.

Cardinal I already have installed, but again nothing screams "insert lyrics here". But at least I can poke at it to see what it does. Have not used it yet, but I had installed a whole bunch of plugins to try out.

[khz]

Yes.
I can't think of a speech synthesizer with a text input field off the top of my head.

[onefang]

Cardinal is a huge complex beastie, and no doubt I'll have fun playing with it. But how is it "Possibly also feasible"? Gimme some clues please.

[onefang]

Ah in the Tags drop down, there is a "Speech" tag, the only one that is greyed out. lol

[khz]

Unfortunately, I can't do that.

[onefang]

" An alien entity hell-bent on mutilating your audio." lol

[onefang]

Cardinal will be lots of fun for a very long time. The closest thing to singing it has is Pink Trombone, basically a mouth simulator, but lots of work to get it to sing actual lyrics.

I'll play with Cardinal later, thanks for pointing it out to me. But I really need to sort out singing soon. Where I type in lyrics, tell it what notes and durations, then get an audio file output.

[artix_linux_user]

There are many many many ways to get your foss system singing:

https://www.sinsy.jp/

https://llllllll.co/t/oddvoices-an-open ... izer/42666

https://en.wikipedia.org/wiki/Utau

quote:

Unlike Vocaloid, UTAU files are not restrictive as it is not a proprietary based license. Therefore, it is possible to use open-source license products with the UTAU software, such as those produced for the Macne series

https://en.wikipedia.org/wiki/Macne_series

love to the world
love to the assholes

[onefang]

Sinsy's source code web site says "Currently, it supports only Japanese." but I want English.

OddVoices source web site says "This project is unstable and in an early stage of development." but at least it says it works under Linux. It sings with an American accent, I want British, Scottish, or Australian.

Utau's Wikipedia site says it only supports Mac and Windows. I did look at it and it's friends before, there's a Linux front end, but as far as I can tell all the voices are Windows, have to be run through WINE, and not open.

Macne's wikipedia says Mac only, which is implied by the name. Again only Japanese.

I'll have a look at OddVoices.

Festival's singing mode is turning into crap. Worked almost OK for the first verse, but the rest get screwed and print errors.

[onefang]

Compiling OddVoices gives pages and pages of errors.

Flinger also doesn't compile on my system.

[tavasti]

Sinsy's source code web site says "Currently, it supports only Japanese." but I want English.

For some reason it seems to be that many vocal synths seem to be japanese or chinese.

[onefang]

I managed to get OddVoices to compile. It sounds as bad as Festival's singing mode, but at least it can be tweaked more, AND I can edit it in MusE since it can grok MIDI files with Lyric events. So I should be able to tweak it to sound a bit better.

I might even be able to write a MusE script to automate the process.

I wonder if I can get Festival to generate the wav file that OddVoices uses to teach itself? Seems feasible, then maybe I can use the higher quality Festival voices that it's singing mode can't use, but with OddVoices.

[onefang]

One of the programs in my package repos is jcadencii, a port of Cadencii to Linux, and the Debian package includes a mini WINE, which I guess is enough to run the Windows VOCALOID, VOCALOID2, UTAU, WORLD, and AquesTone voice generating backends. Can't find any docs about how to actually set up these backends under jcadencii+miniWINE+Linux. Where would I find open versions of these backends?