Create a custom voice for your brand using ElevenLabs

Need to add a professional voice to your content? With Eleven Labs, you can use AI to generate custom voiceover on demand – either based on a ‘real’ voice of you or someone you know, or someone entirely fictional!

In this lesson, we'll cover:

  • Finding the right voice for you:

    • Choosing potential voices from the extensive voice library 

  • Generating a unique synthetic voice with the voice design tool

  • Creating a voice clone of yourself, or someone else (with their permission!)

  • Using the text-to-speech tool to generate great-sounding voiceover

  • Go deeper: other AI voice tools & workflow ideas

Typical marketing use cases:

  • Creating classic audio formats like audiobooks or podcasts

  • Adding narration and voiceover to videos, from short-form TikTok formats to longer YouTube videos or tutorials

  • Offering live audio help, by connecting the voice to a knowledge base, so it can answer customer questions

Why AI can help:

  • Generation at scale: if you have hundreds of articles on your website, you can easily turn them into audio assets at a pace that would be impractical with live voiceover

  • Convenience: if you market an individual or ‘personality’, you can clone their voice, allowing you to quickly create content that sounds like them, without needing to get them back in front of a microphone

  • Distinctiveness and consistency: a tool like this offers a more bespoke sound than the default AI voices offered in some social media apps, while deploying the same voice across all your content creates another element of brand feel, along with your palette, tone of voice and so on

1) Getting started

First of all, we’ll add the voices we want to the ‘My Voices’ section.' Then, we use the text-to-speech tool to generate audio using that voice. Go ahead and login – the free tier will offer you 10 minutes of synthetic speech to play with, while the cheap $5 starter tier offers more credits and lets you experiment with Voice Cloning, too.

2) Finding the perfect voice

2a) Finding a voice in the voice library

To explore the library of original voices, click Voices in the left hand navigation, then Library at the top.

Looking for something specific? You can use the navigation tools at the top of this screen to search for adjectives in the description (like “smooth”, “husky”, “teenager”, “newsreader” or “evil”) or filter by language, accent and suggested use case.

Each accent comes with a demo clip. Once you find a voice you like, click Add + to save it to your Voices for later use.

Tip: speech you generate later will also tend to match the source recording quality, so you’ll want to avoid any recordings that have a tinny sound, reverb, or audio artifacts.

2b) Designing a synthetic voice

You can access this tool from the Home screen, or go to Voices → My voices → Add a new voice, and choose Voice Design.

In this dialogue, you can prompt the AI for the exact voice you want – be descriptive! In the second text box, you can enter the sample dialogue you’d like the candidate outputs to recite.

Clicking ‘generate’ will cook up three possible matches in just a few seconds. If one of them is suitable, click it to save. Not satisfied? Try adding more detail to your prompt, or simply hitting ‘regenerate’ (a given prompt can generate quite a wide range of voices!)

2c) Cloning a real-life voice

If you’re using the Starter tier or above, you’ll have access to Instant Voice Cloning, which can also be accessed by Voices → My voices → Add a new voice. This tool rapidly generates a voice that closely (although not always perfectly!) matches a recording of a speaker. This tool requires as little as 10 seconds of audio, although in our experience, the more you can upload, the better. In the video tutorial, we uploaded 8 minutes of audio to get a pretty good impression. Here’s an example of an original voice vs. a clone:

Tip: The ‘cloned voice’ will match the tone and intonation of the original recording – for example, if you talk in a calm, soft voice, so will the clone. If you intend to create a range of content by the same “person” with different tonal qualities (e.g: chatty and confessional sometimes, straightforward and explanatory other times) then record two+ different sets of audio with them talking in this register and create two+ different voices.

Note: Instant Voice Cloning is probably not quite good enough to ‘fool’ people who are very familiar with the original person’s real voice, but it is a good tool if you want to roughly match the voice of someone – for example, if you want content to sound like it’s VO’d by your founder, but the typical person is not super-aware of what they sound like. If hyper-realism is important, you might want to upgrade to the Professional Tier for access to Professional Voice Cloning – though you will ideally need hours of audio of them talking to get perfect results.

3) Generating speech

Now you’ve added your candidate voices to ‘My Voices’, you can head to the text-to-speech tool in the left-hand navigation. In this simple interface, we add our script to the central pane, while on the right hand side, we can choose the target voice, as well as play around with a few settings.

Boosting ‘similarity’ and ‘style exaggeration’ can often push the output closer to your expectation, while the default settings tend to push the voice to be a bit more “well spoken” or “narrator-y” than you might expect.

All generations use up credits, so while testing voices, it’s best to use just a few sentences to experiment with. When you’re ready to go, you can add your full script for generation all in one go.

You’re now ready to create all the audio content you could dream of, all in your ideal AI voice. Enjoy!

4) Taking things to the next level

asdasdad