I just bought Synthesizer V artificial intelligence voices Solaria (female) and Kevin (male). Amazingly, they can "sing' a midi file if you just type in the words. I am thinking of using it for quick songwriting with BIAB.
Below is a link to my first three attempts to use it with BIAB over the past two days. The first two were disappointing in many ways, but on my third attempt I was able to get something I think is promising.
What's wrong with the first two? I downloaded two BIAB .mgu files from the internet (All the Things You Are and All of Me) and left the chords the way they were, and ran the the melody track through Synthesizer V and bulk copied all the lyrics from the internet. A few places had to be tweaked, but I pretty much let it do it's own thing regarding phrasing and pronunciation. The whole process from start to finish took about 15 to 20 minutes per song.
Although the voices were clear and quite realistic, they just sounded wrong to me (what do you think?) because the timing of the melody was exactly quantized, and no singer would sing that precise on every beat. Also the expression and vibrato was too uniform, and needed some variation.
What's great about the third attempt? So I tried a third song (Crazy) and this time I didn't use the melody midi track, but re-played it on my keyboard without any quantization, trying to be freely expressive. I then added the lyrics and spent more time changing the expression, slides and vibrato in various places. This took me about 2 hours (but I was learning the program, so it can be done faster with experience.) This one sounded a lot more human to me.
All three attempts can be heard here in sequence (just keep listening to hear all the tracks)
Impressive. Where the immersion breaks for me is when they don't take any time to breathe. I'd be keen to see behind the scenes how you put it together. The 3rd one sounds a little scratchy (maybe just clipping), I'd say the first was my favourite. That second song is begging for a duet.
I don’t work with vocals but if I had to, this would be very helpful to me as a composer. I could work with Kevin. Do they have other choices of female vocalist?
EDIT: yes, Eleanor Forte, good only for Anime sounds
I did not hear distortion in the third song, and I did hear your efforts to humanize. Nice work.
Thanks Ryan_R for the advice about adding breaths. You're absolutely right. So I checked the manual and it is super easy to include breaths (just add a "word" spelled "br" and it sings it like a natural breath). I tried this and it instantly makes the track sound more human. It also forces you to leave gaps where a singer would naturally breath, so that is great too. I've only had the program for 3 days, and already I seem to be able to make vocals better than most demos I've heard, including the one you posted. I think the program (Synthesizer V) has gotten better recently and there is a beta version which just came out that has a some new features. I'm still finding more things I can do to improve the performance, so my first three out-of-the-box tests are nowhere as good as what someone should be able to do with more experience.
Great to hear. I listened to a few renditions last night and the most natural sounding one was a Lady Gaga song I think.
The website seemed a bit confusing with paid and free/lite versions of the software and the voice. Many of the stores don't have English translations by default. Which editions did you end up using?
It’s been mentioned above but worth saying again … members on the User Showcase forum have posted songs using synth vocals since their advent. It has been fascinating over the last decade following their development. Some of the latest posts by dcuny, B.D. Thomas and jannesan are well worth the time to listen.
Thanks Ryan_R for the advice about adding breaths...
A simple piece of advice: sing along with your song, and whenever YOU need to breath, so does your virtual singer. And if there's no time to breathe, change the lyrics. No kidding...
I've been playing around with the program (Synthesizer V) for about a week now, and I just realized that you can instantly double a track with a different singer with just one click, to get a very realistic choral sound. Then you can slide the notes around on the doubles and get instant harmonies without having to type in the words again. Here's a small test I did this morning, using 4 different AI voices singing together:
I've been playing around with the program (Synthesizer V) for about a week now, and I just realized that you can instantly double a track with a different singer with just one click, to get a very realistic choral sound. Then you can slide the notes around on the doubles and get instant harmonies without having to type in the words again. Here's a small test I did this morning, using 4 different AI voices singing together:
I think you’ve done well all round with all of these and have made tremendous progress in a short time. “Blue Bayou” is very good, even Asterian sounds real. Usually he sounds (at least to me) like a higher range singer slowed down.
Your tips are very useful. I see you’re on VI-Control, posters on the Synthesiser V thread are really knowledgeable and offer good tips too.
Thanks for the advice about the VI-Control thread, Cyberic. I'm going to check it out now. I didn't know anything about this program until a week ago, but I'm amazed how well-thought-out the functions and interface are. Very intuitive and easy to learn. It's going to take some time for me to really make it sing. But still, it's a blast right out of the box...
At this moment I'm thinking back to the first voice synthesiser I used back in the early seventies. It was a 19" 1U rack unit and all it could do was say the digits zero through nine.
The first time we ran it, it had us in fits of laughter because we were in southern England and as we stepped though the digits, we were quite unprepared for it's distinct southern-states accent ... "fower, faave, ...".
We used them for two or three years until we started digitising our own voice prompts and messages.
Singers A.I. they got me, so I already have Synthesizer V Studio Pro, beautiful, and I have singer A.I. who can do cleaner lowest notes than me, so I'm starting and learning to sing voice bass from Asterian :-D
I'm definitely not a singer and this may be a good tool for me to investigate further. I've found a few decent YouTube videos here and there and may pursue this next time I create a new track.
P.S. - Thanks for the comments on this from all the other forum members too. Lots of good info here.
I recently bought Synthesizer V with “Kevin”, and I am trying to learn how to use it. Can someone explain the “Instant Mode” to me ? I think (I may be wrong) that you can only edit individual notes when you are NOT in “Instant Mode”…. Is that right ? I feel that there must be more to it than that - any pointers would be much appreciated.
Can someone explain the “Instant Mode” to me ? I think (I may be wrong) that you can only edit individual notes when you are NOT in “Instant Mode”…. Is that right ? I feel that there must be more to it than that - any pointers would be much appreciated.
Kelvin, I've only had the program for about a week, but I found I can edit notes in Instant Mode in 3 ways:
1) in the piano roll (next to the words "piano roll") highlight the pencil tool and the squiggly lines (first two) and you will see the pitch and vibrato information, and with your pencil draw whatever you want over the top of them.
2) open the parameters panel (see https://www.youtube.com/watch?v=agSYwd_LQZU) and you have control on the bottom to change anything - select what you want to affect (next to the word "EDIT" and there are eight parameters you can control on a note-by-note basis. This is very powerful and enables you to vary the performance and tone in different places.
3) with the arrow tool (not pencil!) select the note (or notes) you want to change and hit the AI Retakes button (on the upper right below the mic icon) then change the expression amount (lower is often better) and hit retake and it will make a new take which should be different. Play it and see if you like it, or keep making new takes until you get one you like.
I've found that "instant mode" sounds a lot more natural because it adds realistic bends and slides approaching and ending the notes, based on artificial intelligence according to the program's understanding of musical phrasing. Without instant mode you could theoretically make everything yourself, but it would take hours and still be clumsy, so just fiddling with the Retakes and expression amount until you have a good take is a great start. Then you need the two methods above to tweak it more. Instant Mode is just a headstart to get you in the ballpark before manually editing, but sometimes it does an amazing job right out of the box.
There is a lot I don't know, particularly all the scripts you can get, which make things even easier, so I will get into that next, but the best two forums I have found are at:
Thank you very much indeed for that reply Thomas - I will carry on experimenting, using your suggestions and links - I am also over on Vi Control - I wonder whether we need a new thread over there exclusively for “Tips and Tricks”, as the official documentation is not great , but I am not complaining about that, because I am really pleased with the software and I am astounded how cheap it is, for the results it can produce.
To use the Synthesizer V Studio Plugin.vst3 in Biab ...
Musosity, thanks for this. Have you used Synth V as a BIAB plugin yourself? If so, how to you feel about it? I wonder if it is any better than having both programs loaded in standalone. I found that I can drag a track or whole mix directly from BIAB into Synth V (without saving to a file and reloading, etc) so the integration is quite good already, but perhaps there is something better as a plugin?
One advantage, I guess, is that perhaps if you save the BIAB file it will have the Synth V session and information in one BIAB load, but I suspect that it will not save the actual Synth V information, but just load the plugin and you still have to load it? It's easy to drag the BAIB mix into Synth V, which will save it with the voice tracks, and anytime you want to change a chord you can re-drag and replace it in Synth V, so I'm just wondering what your experience might be with it as a BIAB plugin.
I don’t work with vocals but if I had to, this would be very helpful to me as a composer. I could work with Kevin. Do they have other choices of female vocalist?
EDIT: yes, Eleanor Forte, good only for Anime sounds
As of this writing, Dreamtonics says the following "native" English female vocalists exist:
ANRI
Eleanor Forte
Natalie
SOLARIA
Tsurumaki Maki English (giant caveat here)
Weina (caveat here)
However, their definition of "native" is that the program was trained with the singers singing in English, not that the singer is a native English speaker.
It also doesn't help when some of the English songs are put together by non-native English speakers.
IMNSHO, Tsurumaki Maki's English is unusable. She's not a fluent English speaker, and I think it's a mistake by Dreamtonics to list her as an "English" vocal.
In contrast, Weina is not a native English speaker and has a bit of an accent, but I find her voice quite usable... most of the time. Giant caveat: there are some times when she requires creative workarounds on some vowels.
Solaria is the gold standard of native English vocals. Clear, versatile and fun to work with. The personality (and skill) of the voice provider comes through.
Eleanor Forte was the first voice released, and that version over-articulated lyrics. A reworked AI version was released, and that voicebank is much better than the original. There's a "Lite" version of the voice as well, which is free and lower quality.
ANRI has a "young teenage girl" voice - good for that sort of pop sound, I guess. I'm not a fan of that vocal sound, but apparently a lot of people are. Her English is fine.
Natalie is a more recent voicebank, and has a darker, slightly husky voice. It's not as versatile as Solaria, but I like her voice a lot. She has a habit of dropping some consonants from the end of her words.
Depending on what you want, the female trans-lingual voices can be quite good. For example, Qing Su is a Chinese voice I did a song with, and forgot she wasn't a native English voice.
As always, these are just my own highly subjective opinions. Your own ears are the best judge as to whether the voice would work for you.
If you have a finished song you might be best just to drag n drop into Synth V. I tried the vst3 but it was a bit sluggish in Biab. I haven't got into it enough to give it a good try out, I tried it with a vocal line out of some kar files and was impressed how it sounded without tweaking anything else. vocal-test.svp
Astorian is a great new A.I. singer with bass voice color and position and local expert on A.I. singing does not have it in its list (useful for me and David Cuny is truly an expert I respect) yet. For me, who likes singing in his positions, it's unreal - link - https://www.youtube.com/results?search_query=A.I.+Asterian - it was also used for spoken word, but I can't judge that because I can't speak English. That's why I judge his singing. Here is the link to the song - https://www.youtube.com/watch?v=L21ngECYLSw
Astorian is a great new A.I. singer with bass voice color and position and local expert on A.I. singing does not have it in its list...
Sorry, I was just responding to Matt's comment about female singers.
If you go to the SynthesizerV main page and scroll down you'll get a list of all the current singers for SynthesizerV. There are currently only two native English male voices (Kevin and Asterian), but Dreamtonics has said they will soon release a native English male voice oriented towards rock, and Audiologie has hinted they will be releasing a native English voice that will pair with ANRI this quarter, so that might also be a male voice.
In a pinch, you can use a native English female vocalist and adjust the Gender parameter.
Using a non-native male voice in trans-lingual mode would be my last choice. While a number of the female voices actually sound quite nice in trans-lingual mode, the males all seem to retain a heavy accent. Maybe it'll get better in the future.
EDIT: yes, Eleanor Forte, good only for Anime sounds
That's the impression I got from most of the official demos, but it's wrong.
This and this are examples how Eleanor sounds when you chose a vocal range that suits her better. No matter whether you like that sound or not, it's definitely NOT anime
Let me know what you think. I've learned a lot in the past few days, and realize there is a lot under the hood of this program.
(By the way, this is not my arrangement. I just played in the notes of a well-known published arrangement by Gene Puerling, to see how Synth V could handle this sort of thing.)
I listened to some Asterian demos too.... This would be a good voice to use for a Barry White type voice, verrrry deep.
Nice work Thomas, the harmonies turned out awesome! I've been working with the software now too and use Kevin. I'm hoping to replace my own voice with him on an old track I did (It certainly can't be worse than my performance was).
B. D. Thomas, also thank you for your comment. I shouldn't have dismissed it so quickly without trying it; I only listened to their one demo that I could find.
Here is a song that I wrote for my wife, featuring Synthesizer V “Kevin” as the vocalist- I would describe it as “uptempo pop”…. I used guitars from one of the BIAB styles, and the bass and drums and some of the piano were edited from BIAB MIDI styles…
This is a great song Kelvin, but perhaps consider bringing the vocals a little more out in front. Not a criticism, but my personal feeling is that they sit a little bit too far back in the mix, and they needn't. They could 'pop' a little more. Regardless, the song tells an amazing story. Thanks for sharing.
This is a great song Kelvin, but perhaps consider bringing the vocals a little more out in front. Not a criticism, but my personal feeling is that they sit a little bit too far back in the mix, and they needn't. They could 'pop' a little more. Regardless, the song tells an amazing story. Thanks for sharing.
Thank you very much for that feedback - I’ve got to improve my mixing skills, for sure, so I will try to implement what you have said - thanks again !
My goal for working with Synthesizer V Pro is to make a song so I can make a good demo recording and then sing the song with someone or have someone else sing it. I have already created such a demo recording, but in the form of a semi-finished product. The lyrics are in Czech, which I can't use here, which I don't mind, and if so, I'll let them sing, for example, La,la,la,la ... I can't write English lyrics and I wouldn't dare to, because I know what it takes to writing the text in Czech. All instruments are from RealBand and most of them are in midi tracks. I will transfer the created tracks to Studio One and let the virtual background vocals and virtual instruments sound mostly through the Kontakt module. I'm doing the mix and mastering at a friend's studio, which is yet to come. I'm still fine-tuning my singing with an excellent singing teacher, because I need it as an amateur. I have already registered the song with our copyright company OSA. I have the vocals where the text is sung in the sheet music, so I tried out Asterian as my voice (I sing deep in the first part, an octave higher in the second part and I return to a deep position in the third part), the second voice sings most of the verses Kevin and Asterian joins as third voice and Solaria as fourth voice in the choruses. While I agreed with Kevin's second voice, the chorus somehow doesn't fit me and I want to make it more interesting. I'm even looking for advice and feedback. However, I put the finished stuff on SoundCloud, but I don't want to put this there, because it's only in an unfinished state and I don't know if I would be doing the wrong thing if I put it here as a video (not a link) directly in the post. I've seen short videos in some posts in other forums, but they were related to a technical issue.
For me, the usefulness of AI. singers is that if I don't feel like singing a song I've already written (I'm learning to sing it quite slowly, because I'm 72 years old, although I know that many older people can still do it). I am still working here, but the experience is at the beginner level in this area. So far, my edits quite often go the wrong way and then I try to correct it and often go all the way back. But what would I like, I've had it since January 2023 and I'm very happy with my toy. And then it helps me in songs for alto and soprano, or high tenor. And also these A.I. singers can find me places in the parts of the song that can be used for their singing as good ideas. An experienced singer will laugh at this, but I do not have his gift, because I am an amateur singer. I don't worry about it and I also smile and enjoy singing and that's how it should be. Finally, I want to apologize to Dcuny (he is really knowledgeable in this topic - I read his posts on forum.synthesizerv and the fact that he did not list Asterian was not his fault, but mine, because my English (I use Google Translator a lot) did not provide me the right feeling and understanding of his list in the post that it was intended only for a defined group.
I don't know what these crazy new people are doing with these fake singers, but back in my day if we wanted to play a piano we had to get out the John Thompson book and learn it ourselves. And if we didn't practice, our mamas would spank our hands with a ruler until we did learn! Or if we needed a guitar we had to listen to Chet and pick up a six string and learn how to play the damned thing!!
That was before those long haired hippie freaks who couldn't even play came along!
The next thing you are going to say is that you want people to just go to some software program and type in some chords like a bunch of stupid monkeys and make some fake songs with fake instruments and pretend like they know how to play when they don't know nothing!!!
I don't like it and I won't have it here in the US of A. If you want to do something this repugnant why don't you surrender your US Passport and go over there and do it in Canada!!!!!!
I want no part of this fake singing and fake music.
So there!!!
P.S. Wait!
I just went on my computer and done fount out I done already published 165 songs using this robot!
How in the heck did that happen.
Forgive me y'all. I had promised myself to swear off hard liquor this year but it hasn't quite gone the way I wanted it to.
The next thing you are going to say is that you want people to just go to some software program and type in some chords like a bunch of stupid monkeys and make some fake songs with fake instruments and pretend like that know how to play when they don't know nothing!!!
I mean, we are on the BiaB Forums, so .. kind of funny but kind of insulting. I'm choosing to take it as intended humor without the insults. You have to 'know something' to make it work, after all
The next thing you are going to say is that you want people to just go to some software program and type in some chords like a bunch of stupid monkeys
That's actually the least of my worries. You'd at least have to know what a chord is and which progressions work and which don't.
Fast forward 10 years and you'd ask an AI of your choice "please compose a Dire Straits ballad like "Romeo and Juliet", and you'd get a decent song sung by Mark Knopfler. "You: And while you're at it, can you please create a suitable music video synced to the music? AI: No problem...".
And these songs would be "composed" by kids who spend all day wondering why some piano keys are black and some are white.
Originally Posted By: David Snyder
That was before those long haired hippie freaks who couldn't even play came along!
Hey, careful my friend... I'm a short haired hippie freak!
The idea that AI-generated vocals/lyrics/music are not very acceptable for the reasons that David alluded to and other reasons will be a point of view of some people. The idea that human-generated vocals/lyrics/music are not very acceptable to some people is pretty obvious.
I don't personally see "AI-generated" music all that different than "human" generated music. The AI output is generated purely by human ideas. Until the AI becomes able to create its own language and culture, it will always mimic human speech patterns.
The way we have learned to create music has significantly changed in the last thirty years or so. In the past, it was necessary to go see some guitar player live to figure out how to copy what he or she was playing. Now we turn on the computer and search for a tutorial on a particular song.
Don't play guitar? Buy BIAB. Need drums? Well, you get the point.
Being "chained" to the piano and having your hand slapped for failures, both real and imagined, has caused more than a few people to abandon music. Other people have spent countless hours taking lessons to no avail. There are many possible "roads to the destination."
I think we live in a large enough universe of music to allow anyone by any method to create whatever pleases them, even the gangster rap I so seriously dislike.
Like it or not, AI-generated vocals, words, text, art, video, and a hundred other AI-generated things are already in everyday use and will increase rapidly as time passes.
The future of music is unpredictable. What we like and dislike will change. Change is the only thing that is constant.
There are some issues to be sorted out with all this AI business. AI is learning from everything it is exposed to, and what it acquires from some artists' intellectual property is not currently being taken into consideration. Legal challenges will undoubtedly follow.
Billy
EDIT: Long Haired Hippie Freak generated by AI...lol
My long hair just can't cover up my Redneck...lol or so my AI says...lol
I played the guitar for about 50 years, but after shingles, I have a numb finger - the ring finger on my left hand and I lost the motivation to play the guitar because my playing was limited. But much earlier, since January 2005, I have had a Band in a box and later a RealBand, so since then I have been playing music in my head and not in my fingers. I sing in deeper positions and my voice is not as flexible as tenor, alto and soprano. Actually, from the moment I started playing the guitar, when I learned a few chords, I started creating my own songs for the entertainment of a small circle of my acquaintances, instead of focusing on more songs written by someone else. Of course, over time playing in different places like pubs, it happened that I was playing well-known songs with more musicians more people and that night I only played three of my songs, so without much effort after a while I played everything I needed and also songs variously known by other writers, because I liked them. Since I started working with bands, I had a desire to write songs for them as well. But the bands were playing songs that had original lyrics in English, so they wanted me to write the lyrics in Czech. Also, some musician, usually the band leader, let me write lyrics to his melody. Band in a box and even more RealBand replaced my band and that gave me the opportunity to write music for my song and year after year I get more great support in developing music applications. I wrote enough lyrics for different female singers (alto, soprano), so I had the desire to write melodies for them as well, but my singing never (with a few exceptions) gave the demo recording the right charge to convince the singer. And now I'm finally going to write why I've been talking for so long. A.I. singers gives me the opportunity to develop my voice for higher positions and also the inspiration to improve the quality of the melody and maybe finally I will be able to offer a song and I will succeed. And most importantly, I won't have to sing it myself in the studio every time. And Synthesizer V Studio Pro is also great for creating vocals.
The emerging manual for Synthesizer V Studio Pro by the user "claire Regular" is created at the link: https://manual.synthv.info/ I'm already reading it and it's great...
The emerging manual for Synthesizer V Studio Pro by the user "claire Regular" is created at the link: https://manual.synthv.info/ I'm already reading it and it's great...
Thank you for this, the Dreamtonics situation with Synth 5 documentation is frustrating. This looks good. Claire’s manual requires a lot of hierarchy traversal off the little hamburger navigation thingie, though, which makes browsing difficult (I like having an entire manual in a PDF, or a single web page.)
