Like it or not, AI isn’t going anywhere. Ever since ChatGPT first exploded onto the scene in 2022, AI has become an increasingly large part of our everyday lives, and music is no exception.
While AI-assisted music production has been around for a while with services like eMastered , AI voice generators are relatively new. With more popping up every day, it can be daunting to figure out which one will create the perfect voice for your songs.
And here at eMastered we like to be helpful, so we've put together this list of the 5 best vocal AI plugins available right now.
Ready? Let’s go!
What Are AI Vocal Plugins?
For the purposes of this article we're considering AI vocal plugins to be any software, plugin or cloud-based, that creates or manipulates vocal recordings.
Some of them will generate the vocal tracks themselves. Others will use voice changing software to adjust the vocal identity of a recording.
Are AI Vocals Legal?
In terms of the legalities surrounding AI voices it's literally the wild west out there. This confusion stems from the material that the AI technology is trained on, and whether the artists whose voices have gone into the training model consented to such use.
The bottom line is that if you use AI to mimic a celebrity voice, and make it available to the public, you could find yourself in hot water.
Best just keep those funny Ed Sheeran tracks you made between you and your mates.
The 5 Best AI Voice Plugins
Elf Tech
Website
The Low Down
Elf.Tech is an AI voice generator that has only one voice model; Grimes.
Launched in 2023, the open-source software was designed to make the luscious voice of Grimes available to creators and musicians for use in their own music.
Like many of the plugins listed in this article the principal is simple: upload a vocal performance (either your own voice, or someone who has given their permission), and you'll get it returned to you transformed into Grimes' characteristic voice.
There's not a lot of control over how you can manipulate the voice, but then again why would you want to mess with Grimes?
It's limiting, yes. But the beauty of this site is that you have full permission from the songstress to release your song commercially, in exchange for 50% of the master recording royalties.
This presumably means you get to retain 100% of the writer's share, although I believe there's still some issues when it comes to releasing on streaming services that her legal team are working out. If you're into sync music however, you're golden.
Pricing
Elf Tech is completely free to use. The downside is once you create an account there's no (obvious) way of signing out, or deleting your account.
Other Features
The Elf.Tech site also offers a distribution service, and access to the stems for Miss Anthropocene for remixing.
Uberduck
Website
The Low Down
Remember that fake Drake song a couple of years back? It spawned a bunch of similar tracks that used AI voice models to mimic other artists.
One website that made all this possible was Uberduck, a text to speech app that also allowed users to create rap vocals using AI voices that sound like your favorite artist - Kanye West, Drake, and many more were among the custom voices available to use.
The fun was short-lived however, as artists rightfully pushed back against the use of their unique voices without permission.
Uberduck still lives on however, and you can use its 'rap' option to generate spoken word vocals for your creative projects.
You have the option of using a pre-made beat (not good), creating the vocal acapella, or uploading a new beat to rap to (paid version only).
Enter your custom lyrics (or use Uberduck's AI lyric generator based on your prompt), and you'll then be asked to choose your rapper.
There's only 4 options to choose from; a male and a female in both English and Spanish.
There's no option to alter the voices, and while the result is perhaps not chart-topping material it's certainly better than any attempt I could make. Certainly good enough for a place holder to see how the beat is shaping up in context with a vocal line.
Pricing
While there is a try-for-free option on Uberduck, to actually get any real use out of it you'll need a subscription. These start at $4 per month for a non-commercial license (great for mocking up tracks for real rappers to work from) to $60 for the full-on Pro package.
The middle ground 'Creator' package is $10 per month - ideal for the content creator who needs a new voice for their material.
Other Features
Uberduck also offers text-to-speech and image generation, a voice changer application, and a voice cloning option where you can create an original voice based on a single audio clip.
Eclipse Sounds
Website
The Low Down
Based in the US, Eclipse Sounds offers 4 AI voice databases that claim to provide recording studio quality vocals for your track. The downside is that they can only be used in the Synthesizer V DAW.
Each AI model is trained from a real artist's voice, and cover all vocal ranges from bass to mezzo soprano. Each voice comes with a pitch and tempo range as guidelines to work within, and features different 'modes' - read: timbres - that you can mix and match to suit the style and genre of your songs.
Not being a Synthesizer V owner I couldn't check out how the packs work, but I do know you can get them to sing any lyric you want (including for the tenor voice, singing in a different language), and that the performances can be edited as you would any MIDI part in a piano roll style editor.
Based on the demos on their website the results are surprisingly effective. Whether this would translate to the perfect voice for your project is hard to say. But in terms of creating a temp track to send to your real-world singer to work from, the singers from Eclipse could be a game changer.
The artists used to train the AI voice models are paid upfront for their vocal data, and receive royalties on each plugin sold. They can also choose whether to share their real identity, or keep it private. So there's some ethical AI practice going on here.
All the AI voices available through Eclipse are royalty free, and just need crediting as their Avatar on a commercial release.
Like with most sample packs, you can't re-sell anything produced by these models, nor use it to train other AI models.
Pricing
Each voice model from Eclipse ships for $90, with discounts for buying in bulk. Again, note that they only run inside the Synthesizer V DAW .
Other Features
Zip. Nada. Not a sausage.
Audimee
Website:
The Lowdown
Audimee seem to have captured the attention of professional creators as well as amateur dabblers. The company's landing page boasts Nicky Romero, Goodboys, and Hugel among their users.
The demos on the landing page speak for themselves, with some solid representation across many mainstream genres. Yes, even Nickelback fans can make music with virtual rock vocalists!
To convert your own voice into one of the many royalty free voices available on the site you simply upload a clean vocal take, or record directly onto the site.
If you're not a great singer there's options to quickly autotune your performance, or use a built-in pitch editor to do it manually.
From their you choose a voice that suits your vibe. Some of these voices are completely free to use, others require a subscription. For my money there's a wide range of vocal styles and they all sound pretty convincing, even if they lack a certain je ne sais quoi . That's AI for you...
Pricing
The free plan lets you access to 14 AI voices, generate 15 minutes of audio per month, but no commercial use rights. The subscription tiers are reasonable priced at $12, $25, and $49 per month, with the higher tiers offering other voices and more conversion hours per month.
Other Features
The Audimee site also offers users the chance to train an AI model on their own voice, isolate vocals from a mix (careful here - if you use it to train a voice model you'll be on rocky ground), and release acapella covers every week to use in your own remixes .
Lalals
Website
The Low Down
I'm not 100% sold on Lalals by virtue of a) possible ethical violations by offering soundalikes, and b) the other features they offer. More on that later.
However, the demos of AI vocal tracks you hear on the landing page are convincing, if a little saccharine.
The user can filter by category (singer, rapper, celebrity, or politician), language, gender, and whether the AI voice is meant to represent an actual person or is taken from their collection of unique voices.
When you're ready to use the voice changer simply drag an audio file, record yourself, or enter a YouTube link.
There's no option to transform the voice in any way, aside from a basic pitch up or down, for converting a female voice to a male one, or vice versa. Just click on convert and the voice changer will do the rest for you.
The free tier is incredibly slow. High speed processing is available on the paid-for plans, so maybe I'm just being persnickety here. But I'm still waiting for Justin Timberlake to sing Livin' On a Prayer ...
Once it's processed, you can download the resulting vocal tracks in MP3 or high quality WAV format, depending on your subscription tier.
Pricing
With Lalals you can try for free, with a limit of 15 generated minutes per month for personal use only, and low quality rendering.
The paid-for tiers include Basic ($12), Pro ($24), and Studio ($99) which all include more minutes of voice changer use, more voices, and a commercial license.
Other Features
In addition to voice changing capabilities, Lalals paid subscriptions include an option to clone a voice. You can create as many of your own voices as your subscription plan permits, and these can be used for voice changing, or speaking a text input.
The other offering is being able to create music with AI. Bleurgh. Don't do it; stick to producing music yourself.
Other Vocal AI Plugin Contenders
If none of the above voice changer sites suit your fancy, see if any of these sites fire up your creativity:
Considerations When Using AI Voices for Music Production
The rise of AI is a game changer for music production.
As a music producer, using AI voices, or an AI voice changer, allows you to create the perfect voice for your song. It's an incredible tool for the content creator who can sing a tune but isn't a professional vocalist, enabling them to conjure up vocal tracks that match the tone of their content.
Music producers who collaborate with other artists and need guide vocal tracks to work from can also make use of these royalty free apps to boost their creativity.
They're also hella useful for making music just for fun.
But despite their usefulness when producing music, AI vocal tracks come with a certain amount of baggage.
For starters, despite the quality you can get from some of the voice generators listed above, the resulting voices will never match the power or emotion of a real human being singing with all the weight of their emotional experience behind them.
Resources are another factor. AI is incredibly power-hungry, with more energy being used for complex AI generations like speech and singing.
Finally, there's the ethical and legal considerations of what data the voices have been trained on. The apps may claim to be royalty free, but that doesn't always mean you'll be free to distribute the resulting song without any blowback.
For my money, AI voice generators are a useful tool for music producers, but should be used with care, and never as a replacement for a real musician. But that's just my 2 cents.
However you decide to do it, go forth and maketh the music!
