Skip to content

Dataset · parler-tts

Voice text descriptions for LibriTTS-R corpus

parler-tts/libritts-r-filtered-speaker-descriptions

English annotations of speaker and speech characteristics for training controllable text-to-speech (TTS) synthesis.

What's inside

The dataset contains between 100 thousand and 1 million tabular records with English text descriptions. It is based on the filtered LibriTTS-R corpus (about 960 hours of English speech from multiple speakers at a 24 kHz sampling rate). Natural language descriptions of speaker characteristics and pronunciation features are added to the text_description column.

What you can do with it

  • Training TTS models with style and voice timbre control via text prompts
  • Fine-tuning speech synthesis systems to generate voices with given speaker parameters
Tasks
Text → speech
Languages
English
Size
100K–1M rows
Downloads this month
592
Likes
9

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("parler-tts/libritts-r-filtered-speaker-descriptions")