Using Uberduck AI Voice Generator to Create New Music
Ezra Sandzer-Bell- Nov 19, 2023
- 6 min read
Updated: Apr 27
Uberduck is an AI voice generation company that offers text-to-speech features, AI rappers, voice cloning and voice-to-voice conversion.
The website enjoyed viral popularity in 2021 Q4, hosting a small collection of celebrity voice AI models that anyone could use for free. Site visitors made typed in lyrics, generated audio files, and could edit the AI vocals in their DAW. As one might expect, the internet rushed in.
AI voice training became a kind of underground obsession in Discord channels. People were tinkering in Google Colab and Hugging Face, using open source audio tools like Harmonai and SVC to train new vocal models.
As an outsider to that trend, I watched Uberduck come fill the vaccum with a straightforward web app that anyone could use. It offered relatively high quality text-to-speech (TTS) services, with AI models in many different vocal styles.
Uberduck included a small collection of celebrity voices that people could experiment with responsibly. Music producers jumped on the opportunity. YouTube and TikTok creators showed off their AI celebrity workflows, racking up loads of streams.
The viral clip below used the ChatGPT prompt "Write a Drake verse about how he doesn't like beans in his chili" to generate lyrics. Then they added it to Uberduck and made a parody song:
This second video below, from Ed Talenti, similarly played with the AI Drake voice:
Talenti's imitation of Drake, Kendrick Lamar, and Playboi Carti are entertaining more than they are convincing. They wouldn't be mistaken for the actual artist. Youtubers at the time were showing that with the right vocal effects, you could make these ai voices sing.
A controversial AI Drake song made headlines in April 2023, after it drove 600,000 Spotify streams and was shut down by UMG. Uberduck has since removed access to celebrity AI voices, but search engine tools are showing that interest in their product has been sustained globally.
Uberduck's voice-to-voice transfer and voice cloning models put them squarely in line with the other ai voice generators on the market. Their AI rap generation feature also positions them comfortably in the ai lyric-to-song camp.
In this article we'll share the latest Uberduck interface and some of the cool new features it offers. We'll shared some documentation of the classic Uberduck interface as well. Near the end we'll offer suggestions of alternative apps.
Table of contents
Uberduck Classic: The original AI rapper interface
The simplicity of Uberduck's original interface was brilliant. Text-to-speech voice AI was suddenly so easy that the little rubber duck became its perfect symbol.
You could pick from OG rappers like 50 Cent and 2pac, along with newer artists like 21 Savage or XXXTentacion. At the time, Uberduck also included models from famous cartoon voice actors and podcast hosts.
Despite removing celebrity rappers from their AI voice catalog, Uberduck did go on to build an impressive AI rap generator that lines up with any tempo.
How do you generate AI rap songs with Uberduck?
To generate AI rap songs in Uberduck, you'll need to click the AI Rap tab on the left navigation menu and proceed through a three-step process of choosing your beat, generating lyrics, and picking a voice model.
When selecting your beat, you can hit play to audition different tracks or upgrade to the premium plan in order to upload your own music.
During the second step, users choose a topic that they want the song to be about. Uberduck's product team realized that people were hopping over to other services like ChatGPT to generate lyrics, so they built it directly into their system.
If you want to write the lyrics yourself, click the use custom lyrics button in the bottom right corner of the setup wizard. Then you can bypass the "ai magic" step where your lyric topic is turned into a rhyme scheme.
We prompted the AI lyric generator to create a song about being a clever little rubber duck. Maybe Uberduck would come up with something autobiographical.
You can read the lyrics it came up with in the screenshot below - note that they will generate four lines at a time on the free plan.
The third step is where you choose your rapper voice or create your own. As someone who's been with Uberduck from the beginning, this step did make my heart sink a little bit.
There are no recognizable celebrities, with the exception of Grimes (who has publicly given consent for anyone to use her AI voice model). Instead, free plan users find seeral avatars with placeholder avatar images, with names like "Angus (pleading)" that seem to be designed for some other use case.
I personally feel there's a UX blind spot here that Uberduck's team could improve on, by using better model names and images:
Fortunately, when you do pick a rapper and generate the final track, it's a rewarding experience. The AI voices rap on beat and vocally they have a fair amount of depth. Users can see details about the beat, lyrics and voice that they chose. It's easy to share the music on social channels or use the remix option to add a new beat under the vocals.
The AI song generator even comes up with a short custom title for your song and prints it out on the screen while you're waiting for the lyrics to come in. It named our song "CLEVERDUCK":
Uberduck Discord community and TTS API
Uberduck has grown their Discord community to over 24,400 members, up from around 14,000 people in 2022.
Their founder Zach Wener writes and maintains a popular Ubderduck tutorial about building text-to-speech Discord bots. It turns out a lot of people like to use TTS voice overs while they're gaming with friends.
This is a good niche for them. Discord chats are one of the most popular places for voice changing plugins, like Voicemod. Discord mods can use Uberduck's voice clone API to train on samples of their own voice and make a Discord chat bot that speaks and sounds like them!
Using UberDuck with AudioCipher and Autotune
Uberduck doesn't include a melodic AI singing voice option, but that doesn't have to stop you from trying. If you use the text to speech option instead of AI rapper, it will output isolated vocals. A free autotune plugin like GVST's GSnap VST can help you shape the Uberduck vocals into that melody.
AudioCipher continues the text-to-music journey within your DAW. Turn any word or phrase into melodies and chord progressions, in your key signature of choice.
Once you've sketched out a song idea with AudioCipher, weave the melody into the Uberduck audio using autotone and see how it comes together.
Uberduck Pricing and Subscriptions
Uberduck offers a generous free plan that include 300 render credits per month. Their pricing is standard for AI tools, ranging from the $10/mo Creator Plan to a $500/mo Enterprise Plan.
Subscribers to the Creator tier get commercial use licenses, private voices, API access, lower latency servers, AI regenerated saps, and 3600 render credits per month.
The enterprise tier includes some of their best features, like custom voice cloning, templated audio generation, and team collaboration features. Subscribers get dedicated suport and over 500,000 render credits per month.
Alternatives to Uberduck: AI rappers and AI singers
Uberduck offers AI rappers but doesn’t include any melodic singing options yet. Fortunately, there are other AI websites that generate AI music, lyrics, and vocal melodies from text descriptions. We’ve covered them in detail on our text-to-song article.
Readers who want to explore those AI song generators can continue reading through this final section to learn more.
Splash Music & Chirp (AI Rappers and Singers)
Splash Music, Chirp AI, and Riffusion all offer text-to-music generation with AI singing and AI rapping voices.. Chirp and Riffusion are both free currently, while Splash offers some entry level options and a premium access tier.
SynthV by Dreamtronics
SynthV is a vocal synthesizer by Dreamtonics designed specifically for creating music.
The artificial intelligence behind Synthesize V is a deep neural network-based audio generation and synthesis engine. It can create custom voices that sing realistically voices. You can control vocal dynamics like breathiness, nasal tone, and more. As a Japanese company, the app includes a strong focus on anime characters and non-english voices.
Google's Text to Speech API
It should come as no surprise that Google has a text-to-speech API available. They've been busy in 2023 deploying apps like MusicLM and Lyria. I mention the TTS product here as a courtesy for anyone who needs expressive voices for speech rather than music making.
Visit Uberduck's website here: https://www.uberduck.ai/
---------------------
This post was written by Ezra Sandzer-Bell, founder of AudioCipher Technologies.
