WTF? Is that a computer singing? Or an existing vocal snapped to a midi track?
you basically type the words and she sings it..so its computer generated...they did use a real singer to make the vst...as far i understand,...but still you type the words and she sings it
There is a lite version which is free so i downloaded it tonight. The free version is standalone, have to pay for pro version with vst. I just want to see what its about, i am interested.
It took me a while to figure a few things out. I got a voice installed and loaded into a project. I entered some items which are “words”, don’t know yet if they can be phrases. You enter them into basically a midi like editor and edit the words and play it and the voice sings. I believe its a single voice only in free version.
You can import midi and render and output audio.
You can set some characteristics of the voice.
You can set bpm.
There seems to be enough there to get audio back into reaper. I am lazy about writing lyrics, this seems fun to me. This certainly takes more effort than using the vsti within reaper it would seem, i just want to see more what it can do before i spend money.
This is all based on about an hour session of installing and playing with.
There is a lite version which is free so i downloaded it tonight. The free version is standalone, have to pay for pro version with vst. I just want to see what its about, i am interested.
It took me a while to figure a few things out. I got a voice installed and loaded into a project. I entered some items which are “words”, don’t know yet if they can be phrases. You enter them into basically a midi like editor and edit the words and play it and the voice sings. I believe its a single voice only in free version.
You can import midi and render and output audio.
You can set some characteristics of the voice.
You can set bpm.
There seems to be enough there to get audio back into reaper. I am lazy about writing lyrics, this seems fun to me. This certainly takes more effort than using the vsti within reaper it would seem, i just want to see more what it can do before i spend money.
This is all based on about an hour session of installing and playing with.
Dreamtronics states that it uses a hybrid ai/sample concatenation approach.
I am still evaluating the free version and the free (“lite”) voices. The only two that sound pretty good to me and are english are eleanor and solaris. The paid version allows for altering characteristics of the voices and has language translation to english for the chinese and japanese language voices from what i can tell. I assume the non-lite voices are of better quality.
You can render a stereo audio track out of reaper and import it into synth v as a guide for building the vocals. You can render the voices out from synth v and import into reaper as media items. Make sure bpm is set to same in both and vocals line up.
You can do the usual audio “slicing and dicing” in reaper; if refrain are same words every time, only need to create it once in synth v.
Some of this effort would be different with the paid version as i assume working with it as a vsti in reaper is the same as any other vsti.
My plan is to use these voices as backup singers in my songs; maybe make my singing sound better, haha!
Looks like about $80 for the paid version and about $70 per voice if i understand the pricing. At some point i will purchase; i’ve heard enough to be interested. It’s fun to use and is inspiring me to work harder on lyrics.
The editor is a little difficult to move around it; but i have only been at it for an hour or two with reading any documentation as of yet.
Nice use of technologies..sounds pretty good here.
The japanese have been at this for years- yamaha has been developing vocaloid and it is big with the manga/cosplay crowds. https://www.vocaloid.com/en/
Considering the level of machine learning available now- "real musicians" are actually getting a run for their physical money.
Seems exciting but also a bit worrying at the same time?
I suppose people might have to make these choices eventually being all that is offered,while they may be giving away their physicality in favour of advancing technologies,or not.
Might make a pretty boring concert or gig,who knows just yet..not been to a virtual gig like that,not sure i want to either.
Obviously perfect for exploring potentials.
In my opinion this doesn’t replace real singers just as drum programming doesn’t replace a real drummer, or bass programming doesn’t replace a real bass player; however, context of the song and the creative desires of the artist matter.
I see it as another tool to use in the songwriting process.
I use the paid version of Synth V Studio and Eleanor Forte AI, and with a bit of work it can sound surprisingly good - in a spaced out kind of way.
If you just feed it midi and words it'll sound like the worst autotune junk you can imagine, but once you play with the timing and articulation a bit, and sit it in a mix, it can be pretty nice.
I use the paid version of Synth V Studio and Eleanor Forte AI, and with a bit of work it can sound surprisingly good - in a spaced out kind of way.
If you just feed it midi and words it'll sound like the worst autotune junk you can imagine, but once you play with the timing and articulation a bit, and sit it in a mix, it can be pretty nice.
How much work (time) does it involve to make it sound as realistic as the demos? Is an acappela convincing enough or it has to be masked in a mix?
To me it's kind of sad... Talent being (further) replaced by technology.
Quote:
How much work (time) does it involve to make it sound as realistic as the demos?
I read something once...
"A drummer can play a 3-minute song in 3-minutes. How long does it take you to program the drums?"
Kinda' true and kinda' funny! But, it does make a point.
I also read something once about a dance remix of some of a pop star's recordings. The guy was spending LOTS of time re-tuning and adjusting the tempo to make the songs blend together. Mostly, trying to make the re-tuned vocals sound realistic. And I was thinking, why not just bring the dang singer back-in? ...She's not dead! Or since the whole thing is pretty-much fake anyway, bring-in someone to mimic her voice if she's "too busy".
Synthesizer-V is the most exciting new VST I've played with in years! I bought the pro version along with the Ken and Solaria voices and it's absolutely amazing how good it sounds.
Check out the 52-page thread on the VI-Control website for LOTS of great examples (and tutorials):
Synthesizer-V is the most exciting new VST I've played with in years! I bought the pro version along with the Ken and Solaria voices and it's absolutely amazing how good it sounds.
Check out the 52-page thread on the VI-Control website for LOTS of great examples (and tutorials):
"A drummer can play a 3-minute song in 3-minutes. How long does it take you to program the drums?"
I hate that saying.
If you're programming drums because you can't play drums, but you also can't program drums very well, then yes ... it will take a while to program the drums. But how long would it take you to learn to play the drums so that you don't have to program them? How long did it take the drummer in that saying to learn to play drums? (probably longer than it would take you to program the song)
Take a mediocre drummer (or worse), and see how long it takes them to record a good, mistake-free take of that 3-minute song (or how long it takes to record several mistake-ridden takes and edit them into one mistake-free take).
Then take someone who's been programming drum parts for 20 years (and who probably has a great library of drum riffs they can drag/drop into the DAW), and see how long it takes them to program the same 3-minute song.
I bet the drum programmer wins.
In fact, even if we use an experienced drummer, they are still limited to playing live, and it will take them 3 minutes to record the 3-minute song (plus maybe some rehearsal time if it's a difficult song they've never played before).
If we allow the drum programmer to drag/drop drum riffs from their library, they could be done in less then 3 minutes (drag in a 2-bar intro riff, drag in a 16-bar verse riff, drag in an 8-bar chorus riff, ... etc. ... done!). Come one, drummer ... what's taking you so long to play that song?!?
I admit ... if we asked Neil Peart (RIP) to jam for 3 minutes, it would take an experienced drum programmer a LOT longer than 3 minutes to program the same performance. But that's not quite the same thing ...and also, how long did it take Neil to become the great drummer that he was? (probably more than 3 minutes)
__________________
Best Regards, Ernie "lunker" Lundqvist
BDSM (Bad Dog Studio Musicians)
Windows 10 running on Z390 + i7-8700
Anyone got any ideas for setting up a feedback loop between some of the online app versions of these AI services? I envision the robots talking to each other and the system running off the deep end in a closed loop.
I programmed my synthesizer to make a nice string ensemble patch. Then I play it into Reaper, correct the Midi timing, and velocities and then render it to a nice audio track. What a waste of time, as I could have just hired a real string section all along. I’m sure it takes them 3 minutes to play a 3 minute piece too.
I also read something once about a dance remix of some of a pop star's recordings. The guy was spending LOTS of time re-tuning and adjusting the tempo to make the songs blend together. Mostly, trying to make the re-tuned vocals sound realistic. And I was thinking, why not just bring the dang singer back-in? ...She's not dead! Or since the whole thing is pretty-much fake anyway, bring-in someone to mimic her voice if she's "too busy".
Agreed; there seems to be a high number of pop stars out there who are basically just dancing sex symbols who can’t sing very well, all just to sell sh-t to young women who are generally a very gullible population and easily shamed into buying sh-t they don’t need. And yes, I feel like tuning vocals has gone too far in the industry; it removes emotion out of the voice. I only tune individual notes if they are way off, for example if the singer had a bad vocal day and missed a note or two, but that’s it.
This just sounds so much like when people in the 70s and 80s would complain that synthesizers 'play the music for you'. 30 years from now, no one will think anything of autotune, it'll just be another tool for music creation. Complaining about autotune is like complaining about musicians playing the piano rather than singing.
I suspect that eventually Autotune will get to the point where it can tune vocals while preserving all the other little nuances by changing those parts with the tuning. Melodyne is already pretty close to being there if used with a light hand.
Sometimes you want to be able to hear the autotune, because it's being used as an effect. Complaining about audible autotune is like complaining about audible reverb. Why not sing in a real church rather than use digital reverb to emulate a church-like space? To claim one is valid and the other is not is counter to everything that art is about.
As for inaudible autotune, this is already possible, it just requires skill, people who complain that autotune is always audible are just giving away the fact that they aren't competent enough to use it properly or achieve the results they're after.
I just started to work with Synthesizer V lite (free version) and the new 1.7 version version is very easy to use:
- record a melody in Reaper with a piano sound,
- import the MIDI file into SynthV,
- choose Solaria voice,
- type lyrics,
- render the result into WAV file,
- import the WAV file into Reaper to make the arrangment.
The hardest part is to find the lyrics because of rhymes, rythme and my english skill.
Many options are available in SynthV but I don't need them for now.
I was using sample-based VSTi until now for the vocal track but I was not able to write english lyrics for a pop song. I tested EW and Realitone word builders but they are still limited. The other software are restricted to opera or etheral voices.
Some Synthesizer V competitors:
- Emvoice,
- Plogue Alter/Ego: free and real-time VSTi),
- Vocaloid,
- UTAU/OpenUTAU.
Is there any convincing AI around this available to play with?
I recorded a punk band and one of their songs would make a great alternate version with Bill Cosby singing it from 1st person. (I could overdub and replace a few "he's" with "I" myself.)
Is there anything close to a toolset to put something like that together or is it still a deep dive into AI training and tech reading that makes your head start spinning after many hours?
I can stumble across a few websites with "celebrity deepfake" bots for a handful of people and hear what I typed in warble/chirped back. (Couldn't find Cosby though. That would have probably been good enough too.)
How many python module installs and command line madness is needed for any of this nowadays?
Is there any convincing AI around this available to play with?
I tested some AI voice cloning but I found nothing good for singer cloning. I did not look deeply.
I'm pretty sure you can find something but I think AI programming skill is required.
Is there any convincing AI around this available to play with?
I recorded a punk band and one of their songs would make a great alternate version with Bill Cosby singing it from 1st person. (I could overdub and replace a few "he's" with "I" myself.)
Is there anything close to a toolset to put something like that together or is it still a deep dive into AI training and tech reading that makes your head start spinning after many hours?
I can stumble across a few websites with "celebrity deepfake" bots for a handful of people and hear what I typed in warble/chirped back. (Couldn't find Cosby though. That would have probably been good enough too.)
How many python module installs and command line madness is needed for any of this nowadays?
There is a "toolkit" available, it's called NNSVS. I'm developing singing databases with it too and its gives some really nice results. It's also being updated constantly.
I would say it would have fooled me at a crowded bar.
Also, so useful to perhaps give some instructions for a human singer to indicate what you are looking for and probably great to fill in for a shady performance or make a chorus track.
Using the links I found on this thread I installed the Synthesizer V program and the Solaria voice. I then spent a good hour figuring out how to get it to not sound like a chipmunk. By the next day I had what sounded like two pretty girl backup harmony voices plugged into a song I'd been working on, and they sound pretty darned good as long as they're backup and not stepping out in front.
Some steps I suggest (for two harmony singers to accompany you).
1) In Reaper, setup two piano tracks and play the MIDI notes you want for the harmony. Pay particular attention to the starting point of the notes. You'll fix the end points in Synth V.
2) Ensure Synth V is set to the same BPM as the Reaper project.
3) Import your own vocal WAV into Synth V.
4) Export the two piano MIDI parts to MID files and import into two tracks in Synth V. Align them with your vocals and work on note lengths, articulations, lyrics, etc. If it's just a repeatable refrain you need to do then you only need to it once.
*Work on the way the words sound more than the timing, because once you export them from Synth V to WAVs and import to Reaper, you can use stretch markers to perfect them.
I have to say, as long as the harmony vocals stay behind the lead vocals, you could fool nearly anyone.
We were completely blown away by the sound and articulation possibilities of Solaria. They did a great job and I'm excited to see what voices will follow.
How about a voice bank from The Weeknd? How exciting to live in these times...
I'd like to try and break it one day, see how it reacts to nonsense words and glitch effects, randomised MIDI with probability percentages for instance. Could sound fantastic. The first sounds that pop in my head are the tape voice splicing from Eno/Byrne's My Life in the Bush of Ghosts. It could be give some real interesting effects, otherworldly voices, happy accidents a go go.
And yeah, as soon as I saw the thread title, I wondered how long it would be until boomer town showed up I can sing, I've hired singers, and I would again, but it doesnt mean this isnt fascinating to me and highly useful to some and inevitable in its existence.
Better to light a candle than to curse the darkness.