Skip to main content

​
Introduction

BTTS v2 allows you to create custom voice clones from short audio samples. This guide covers how to create high-quality voice clones and format your text inputs for the best results. To access the voice studio, go to: https://app.bland.ai/dashboard/voices?tab=studio

​
Creating a Voice Clone

To create a new voice clone:
  1. Navigate to the Voices Studio tab
  2. Click New Voice
  3. Upload a 10–15 second audio sample

​
Sample Requirements

For best results, your audio sample should have:
  • Clear enunciation
  • No background noise
  • A single speaker
  • At least 100ms of silence at the beginning and end (avoid cutting off mid-word)
  • Emotional variety (for more expressive output)
Tip: You can use Audacity to edit and trim your clips. To download audio from YouTube videos, we recommend yt-dlp.

​
Configuration Settings

BTTS v2 provides two configuration options to fine-tune your voice clone.

​
Speed vs. Smoothness

This setting controls how many frames the system waits for before decoding the first audio frame.
  • Lower values: Faster initial audio output, but reduced quality in the first few hundred milliseconds
  • Higher values: Smoother output, but potentially slower response time

​
Clone Consistency Boost

This feature adds additional prompting to align output more closely with your original voice sample.
Important Note: Disable Clone Consistency Boost if your agent needs to code-switch or speak multiple languages not present in your sample. If you’re experiencing broken or distorted outputs, try toggling this setting off.

​
Text Formatting

​
Spacing Requirements

BTTS v2 processes text by splitting on spaces. Each word is generated individually, so proper spacing is essential for accurate output. Correct formatting: 
Hello <vocal burst> world.
Incorrect formatting: 
Hello<vocal burst>world.
Hyphenated or concatenated words may cause reduced accuracy: 
# Avoid
concatenated-word
from-the-same-array

# Preferred
concatenated word
from the same array
Note: For Asian languages (Chinese, Japanese), the tokenizer handles character segmentation automatically—no manual spacing required.

​
Text Normalization

The model automatically normalizes common text patterns before processing. You do not need to pre-format these inputs:
  • Emails: [email protected] → john @ gmail dot com
  • URLs: www.google.com → www dot google dot com
  • Smart quotes: Converted to straight quotes
  • Repeated punctuation: Hello!!! → Hello!
  • Missing spaces: End.Start → End. Start

​
Working with Numbers

For best results when generating numbers:
  • Keep number sequences to 6 characters or fewer
  • Format numbers as words when possible
  • Always end sentences containing numbers with a period
  • For long number strings, use pause tags to break up the generation
# Recommended
One two three four five.

# For long sequences, add pauses
One two three <|0.5|> four five six.
Important Note: If your sentence does not end with proper punctuation, the model may drop the last number.

​
Using Punctuation

Punctuation directly affects speech delivery in BTTS v2:
  • Ellipses (…): Adds pauses and weight
  • CAPITALIZATION: Increases emphasis
  • Standard punctuation (. , !): Provides natural speech rhythm
  • Dashes (—): Can add stutters or breaks

​
Best Practices

  • Choose a voice sample with emotional variety for more expressive output
  • Use natural speech patterns and proper punctuation
  • Avoid unusual formatting (excessive dashes or ellipses) for timing control
  • Always include proper spacing between words and tags
  • End sentences with appropriate punctuation, especially when numbers are involved