Hi everyone,
I should firstly mention that I’m using Cubase 5 and Windows 7 Ultimate, in case that has any bearing on this ( which I feel it might have ).
I’ve been using both of the above for twenty-odd years and am well versed in the techniques of tuning an out-of-tune vocal track ( singing is not among my talents ). However, I recently heard that software now exists which basically does the equivalent of taking your lyrics and melody and giving them to a professional singer to sing. Like many things, it may be that this has been around a while and I just didn’t know of it; but I would like to hear people’s thoughts and experiences. Is there any particular software that would be recommended? - Why? - Would it work with my system, or does it work completely independently of that? - Can you make your voice sound like any singer ( e.g., if I wanted to - which I don’t, particularly - but could I make my voice sound like, say, Robert Plant? - as a male, could I make my voice sound like a female singer? )? These are the kind of questions I’m interested in. I would be grateful for any thoughts, opinions, experience or advice in this regard.
Thank you.
This guy has a bunch of vids on AI vocals. This is just one of them.
I think even the best current technology results in at least a little bit of a synthetic vocal quality, but it’s all getting a lot better.
Also, you can hire live real singers on line, I’ve done that with Studiopros in the past, I think fiverr may have some, other sites as well.
A magic wand 
I’ve done this before and liked the results a bunch. It’s not just singers either, they’ve got all kinds of players. Lotta burned out LA musicians might appreciate a gig right now.
I’ve off & on been playing around with synthetic voices since early Vocaloid days. Until recently the ratio between sounding OK and being frustrating to use has not been what you’d want. Stuff like a word/syllable sounding fine on a quarter note but totally weird on a dotted half.
Just this week I’ve been using (and learning) Emvoice for a Project and it is pretty straightforward to use. Initially the vocals sound a bit robotic, but selectively adding bits of vibrato makes a huge difference. Soloed lots of it sounds pretty realistic with occasional ‘not quites’. But with backing music and vocal processing like reverb & delay it sounds plausible.
All these new vocal synths work differently and therefore consume different resources, so it’s gonna depend on specifics. For example Emvoice is a pretty low resource user because most of it’s work is locating & downloading little bits of Audio. The decision about which exact bits of Audio to use is determined by AI on their server. They do have a trial available.
You’ve got a pretty ancient rig so I’d expect in general lots of modern software might have issues. FYI, you are aware of the upcoming retirement of the old Steinberg licensing system?
Let’s ask an AI
Summarize the latest generation of vocal synths noting their strengths & weaknesses?
The latest generation of vocal synthesizers has made significant strides in realism and versatility. Here’s a summary of their strengths and weaknesses:
Strengths:
- Realism: Modern vocal synths like Synthesizer V and VOCALOID 6 offer highly realistic vocal sounds, making them suitable for professional music production.
- Multilingual Support: Many vocal synths now support multiple languages, allowing for diverse vocal performances.
- User-Friendly Interfaces: Tools like Emvoice and Sinsy feature intuitive interfaces, making them accessible to both beginners and experienced users.
- Integration with DAWs: These synths can be easily integrated into digital audio workstations (DAWs), streamlining the music production process.
- Customization: Users can manipulate various vocal characteristics, such as pitch and tone, to tailor the output to their specific needs.
Weaknesses:
- Cost: High-quality voice banks and advanced features can be expensive, making it a significant investment for some users.
- Complexity: Fine-tuning certain voice banks and achieving the desired sound can be challenging and time-consuming.
- Limited Voice Styles: Some vocal synths have a limited range of vocal styles, which may not suit all musical genres.
- Internet Dependency: Cloud-based synths like Emvoice require an internet connection, which can be a limitation in some situations.
- Learning Curve: Mastering the full potential of these tools can take time and practice, especially for those new to vocal synthesis.
Overall, the latest generation of vocal synthesizers offers powerful tools for creative music production, but they come with a learning curve and potential costs.
Can you provide a list of the synths along with their specific strengths & weaknesses?
Sure! Here’s a list of some of the latest generation vocal synthesizers, along with their specific strengths and weaknesses:
Vocal Synth Strengths Weaknesses Synthesizer V (SynthV) Realistic vocal options, supports multiple languages, free vocal modes, regular updates Voice banks can be pricey, some female voice banks might sound similar, fine-tuning can be challenging VOCALOID 6 High-quality voice banks, flexible and customizable, improved natural-sounding voices Expensive, fewer English voice banks available, limits on multilingual singing Emvoice Realistic voices through granular synthesis, cloud-based, user-friendly plugin Limited vocal style options, requires an internet connection, no multi-language dictionary Sinsy Advanced features, handles various languages, intuitive interface, customizable voices Limited advanced features, some users desire enhanced pitch correction or real-time editing Each of these vocal synthesizers offers unique features and capabilities, so the best choice depends on your specific needs and preferences.
Thank you to everyone for all the information ( even the ‘magic wand’ suggestion, which in my case probably isn’t too far from the truth ). I’m in the process of looking into the various suggested resources. I note that cost seems to be a significant ( if not surprising ) issue, although I haven’t yet got as far as finding out specifically how much the various options do cost. Learning curve is certainly also a consideration; even after so many years, I wouldn’t honestly say I’ve fully mastered Cubase yet, let alone a whole new learning process. Given that I do this almost entirely for personal pleasure as opposed to any kind of commercial or financial gain, I think probably something mid-range is likely to be the option of choice, if in fact this does turn out to be a workable course of action.
One of the things I’ve noticed ( so far, although I haven’t had time to look at the options in greater depth ) is that some of the voices do seem to sound quite similar to one another, and all of the ones I’ve heard seem to be aimed primarily at the younger, chart-friendly types of music rather than ( I’m not ashamed to admit it …! ) the older prog rock / theatrical kind of thing. Also, most of the voices I’ve heard so far are American - not that I’ve got anything against that, I hasten to add and emphasise!! - but it doesn’t help if you’re looking for a more English-sounding voice.
Still lots of homework to be done. If anyone has anything to add, I’m still very interested in any thoughts, comments or advice. Does anyone have any first-hand experience of the sort of thing we’re talking about?
By the by: I hadn’t considered the Fiverr option and will do so; my only reservation is that having, on occasion, advertised for singers and had a very wide variety of results, in more ways than one, I don’t know if I could be bothered to go through the possible grief of explaining exactly what I was looking for and then ending up with something completely different. I may be being unfair, and if so I apologise; I don’t have any actual experience of anyone from Fiverr, so it may be that I’m being unduly pessimistic.
Same situation here, long retired ex-pro still making music for personal enjoyment and a few streams. We (wifey & me) used to sing lead and backing vocs but now our voices have aged so we always look for a session singer when we need vocals.
I agree that most ai voices are younger sounding although Sonarworks AI have some older models. I also find that while I can get a half decent male ai to improve my vocal sound, male to female is still a challenge. What I quite like about Sonarworks is the fact you need to have a decent vocal take first because the ai takes the characteristics of your voice - intonation, accent, nuances, sound and adjusts whichever voice model you choose. Sing in a cut glass English accent and that’s what you get out of it. It won’t magically give you a perfect vocal from rubbish input.
I have also used fiverr and while I’ve generally been happy with the vocalists’ voice quality, it’s the technical quality of the vocal takes that can be disappointing. Mostly that’s down to a session singer’s choice of mic, or there’s room ambience as not all have vocal booths. And sometimes despite asking for completely clean non treated vocals I have ended up with something that has obviously been processed. In some cases I think that’s to do with removing background noise. I’ve more or less come to the conclusion that it’s better to find someone locally and get them into my little studio here with accoustic treatment and vocal booth and a decent mic. And there’s the bonus of being able to give some instant feedback and get the vocal take I want.
Good luck in your quest Firefox7 and thanks for starting this interesting thread.
Just by way of a ( hopefully ) amusing anecdote ( and not that anyone asked, to be fair! ), allow me to elaborate upon why I’m iffy about recruiting singers, regardless of the source. I don’t know whether our American friends get a British sit-com called ‘Only Fools And Horses’; those of you who are familiar with that programme may possibly sense where this is going, but I swear this story is true.
Twenty-odd years ago, I recorded a cover of ‘Every Breath You Take’ ( with no disrespect intended, I’m assuming there’s no-one in the civilised world who won’t be at least somewhat familiar with that song ). I put an ad in the paper for singers, male or female. I wasn’t in a position to be able to make any kind of payment, so the deal was that if they didn’t mind recording the vocals for free, I would give them a copy of the finished result which they would be free to use on their own demos if they so chose.
There were quite a few replies, and I asked people to send me their demos. People did, and eventually I narrowed it down to about five or six contenders. One of them seemed to have a particularly good voice, quite well suited to the song, I thought, and we agreed that he would come along and give it a go. Suffice it to say, he had chosen his demo songs carefully.
He certainly lacked nothing in willingness: he’d travelled about twenty miles across London to get to where I lived at that time, which, considering there was no payment involved, seemed like quite a shlep. Anyway, we got set up, he had a bit of a warm-up, and there I was ready to hit the record button.
You have to bear in mind at this point that this was not ‘Britain’s Got Talent’, and I am not, thankfully, Simon Cowell. With him or any of the other various singers who had helped me out on recordings I was doing at that time, in light of the fact that he / they had made the effort, I always felt the very least I could do was to see it through and do my best to produce the most professional result that was within my capability and my technical capacity ( although admittedly that is in my nature anyway ). It was around ten a.m., and there was potentially a long day’s work ahead. With optimism in my heart and plenty of coffee in my cup, I hit the record button.
Many things have surprised me in my life. My beautiful wife-to-be telling me that she would marry me, and then some time later telling me that she was pregnant; my daughter telling me that she was pregnant and that I would therefore end up a grandfather before the age of forty; Boris Johnson becoming prime minister. None of these events, nor any other that I can recall, left me with the same look on my face or near-volcanic feeling in my stomach as I had when I heard the words:
“Evwy bweath you take …”
It was a very very long day.
Witty anecdotes aside …
Sonarworks sounds like an interesting possibility, although as with some of the others I’ve looked at, I’m concerned as to whether it would work with my existing system ( Windows 7 Ultimate / Cubase 5.1 ). This is obviously a significant consideration, as I’m getting along perfectly fine with the system as it is, and have no wish to get involved in costly, laborious and from my point of view unnecessary updates. If it will in fact work with my system, I might give the free trial a go and see how that pans out.
Meanwhile, still looking at the various other possibilities. Any other thoughts, comments and suggestions still welcome ( clean ones preferred ).
I’ve heard rumors of, but can’t confirm, that some AI sites lay claim to at least partial copyright of anything you send in, maybe want to look at that first if you’re thinking of going that route.
Please post back if you learn anything about that!
I will certainly do so, although if that does turn out to be the case, they can stick that where the sun don’t shine. Whilst I have no expectation of ever making any money out of my music, at the same time, I’m not about to give the legal rights to it away either!
She has a lot to say about it, I don’t know her credentials or level of expertise
I checked the FAQs on Sonarwork’s website. Looks good there regarding copyright - please check on your own there.
five characters
I’ve also checked Sonarworks’ End User Agreement, and I can find nothing in there that relates to the copyright of any material processed using their software, only to the software itself and anything that might come under the heading of their ‘intellectual property’, which is fair enough.
On the face of it, Sonarworks seemed like as good a place to start as any, particularly since they offer a week’s free trial. Unfortunately their software isn’t compatible with anything earlier than Windows 10. I gave it a try just on the offchance, but not surprisingly, it wouldn’t install.
My investigations continue. Offhand, does anyone know of any resources that will run with Windows 7 ( and Cubase 5.1, where Cubase needs to be involved )?
No, sorry.
Since both products in those versions are out of support, I assume current vendors don’t target it anymore.
I can’t deny that does seem a reasonable assumption. If anyone has any thoughts as to how I might solve this problem ( i.e. the voice AI problem, not the out-of-date software problem ), by all means send 'em this way.
Have you had a chance to take a detailed look at some of the AI options in the Lanewood Studio vids (low cost or even free) or the online singer for hire options (higher cost) yet?
Initially I kept Audimee ( the resource recommended by the Lanewood Studios guy ) in reserve, as I was hoping to find something that I could use on my own computer ( which would still be preferable if it were an option ). However, in light of my experience with Sonarworks, and in response to Alexis’ last post, I thought I would give Audimee a go and just see how it worked out.
My initial impression is very good indeed. I tried two different vocal tracks, one sung by myself and one by a friend and ex-bandmate. In both cases - since the amount of conversion time available in the free option is limited - I tried a male-to-female voice conversion as I was interested to see how this would come out, and in both cases it was a resounding success ( at least to the extent of my own expectations; obviously everyone will have their own ).
My only reservations so far are:
Firstly, as far as I can tell, the only option apart from the ( understandably limited ) free one, is to pay by subscription. I understand the reasoning behind this; however, I don’t know whether my need of this particular process is sufficiently regular to justify paying for it every month whether I use it or not ( and there will undoubtedly be months when I don’t, probably more than those when I do ).
Secondly, even with the paid options ( not sure about the dearest one, without checking it again, but I wouldn’t be likely to go for that in any case ), if you want to ‘try out’ various voices for one of your tracks before actually committing to a specific one, I’m not sure if all of that counts toward your total available ‘conversion time’ for the month.
All of that being said, so far I’ve spent no more than perhaps an hour or so working with it, which is clearly not very long at all. In view of the fact that the results were better than I’d expected them to be, I’m optimistic.
If anyone has any comments on any of the above, they would be welcome as always; alternatively, if anyone has any other suggestions to offer ( bearing in mind my not-exactly-brand-new system ), they too would be welcome.
In light of the fact that this gives you far more control than even the best ‘online singer for hire’ option, for the time being at least, this is the direction I’m looking in.