ElevenLabs Best Voice Settings Tutorial For Beginners (AI Voice Generator)

Education

Introduction

In this AI voice tutorial for beginners, I'm going to show you the best ElevenLabs settings for creating voices. We'll be using the free version of ElevenLabs, which allows you to generate speech with some limitations on character count. When you sign up, you get a quota displayed on the homepage showing how many characters you have left for free use.

Creating Your First Sentence

First, create a sentence, such as "Please subscribe to Mr. Money." While generating a voice with this sentence, you can select from various pre-built AI voices available in the dropdown menu. Each voice has its own unique profile and style, including some that sound like celebrities. You can sample these voices by pressing the play button next to their name.

Customizing AI Voices

After selecting your desired voice, there are several settings you can adjust to fine-tune your voice output:

  1. AI Model: The most recent and advanced model is the "11 Multilingual Version 2." It supports multiple languages and offers the best performance.
  2. Stability: This controls how stable the voice is. Lower values provide more emotional range but can result in randomness, while higher values may make the voice sound monotonous.
  3. Similarity: This dictates how closely the AI mimics the original voice. Higher similarity can reproduce artifacts if the original recording has poor quality.
  4. Style Exaggeration: This makes the voice more exaggerated in its style of speaking. However, it's generally advised to keep this setting at zero unless specifically needed.
  5. Speaker Boost: It improves the similarity to the original speaker but increases latency.

Practical Example

Create a simple sentence like "Please subscribe to Mr. Money on YouTube now." Add emotional cues such as "he shouted angrily" to explore emotional expressions. However, remember that these cues will also be read out by the AI. Another way to add pauses is by using three dots ... or commas , to help the AI understand where to pause naturally.

Adjusting Settings

We found that the best settings for a natural-sounding voice are:

  • Stability: 70%
  • Similarity: 60%
  • Style Exaggeration: Set this to around 15-18% for slight emotion but avoid instability.

If you want to add more emphasis or different tones, you can tweak the settings accordingly. Use commas for brief pauses and capitalize words to add emphasis such as "Mr. Money."

Conclusion

These settings make a significant difference in improving the quality and naturalness of the AI-generated voice. After generating your final voice, you can download the audio by clicking the download button next to the play bar. This file can be further edited using tools like Audacity for more refined customization.

Keywords

  • ElevenLabs
  • AI Voice Generator
  • Voice Settings
  • Customizing AI Voices
  • Stability
  • Similarity
  • Style Exaggeration
  • Speaker Boost
  • Emotional Expressions
  • Pauses

FAQ

Q: What is the best AI model to use in ElevenLabs? A: The best model currently available is the "11 Multilingual Version 2," which supports multiple languages and offers the best performance.

Q: How do I add pauses in my generated speech? A: You can add pauses by using three dots ... or commas , within your sentence to instruct the AI where to pause naturally.

Q: What should the stability setting be for natural-sounding voices? A: A stability setting around 70% strikes a good balance between naturalness and emotional range.

Q: How does the similarity setting affect my voice? A: The similarity setting dictates how closely the AI mimics the original voice. Higher similarity can reproduce artifacts if the original recording has poor quality.

Q: Can I add emotional tones to my voice? A: Yes, you can. However, it’s best to write your sentences as if they are in a book, using phrases like "he shouted angrily." Keep in mind these phrases will also be read by the AI.

Q: How do I download my generated voice? A: You can download your generated voice by clicking the download button next to the play bar on the ElevenLabs website.