Skip to content

Dataset · Matthijs

CMU ARCTIC speaker voice embeddings

Matthijs/cmu-arctic-xvectors

7,931 voice embeddings (X-vectors) for 7 speakers, extracted from the CMU ARCTIC speech dataset.

What's inside

Contains 7,931 .npy files, with one 512-dimensional vector (X-vector) per utterance from the CMU ARCTIC corpus. The embeddings were extracted using the speechbrain/spkrec-xvect-voxceleb model for seven speakers with different English accents (American, Canadian, Scottish, and Indian).

What you can do with it

  • Training and fine-tuning multi-speaker text-to-speech (TTS) models
  • Voice cloning and audio-to-audio voice conversion tasks
Tasks
Text → speech, Audio → audio
Size
1K–10K rows
Downloads this month
25.4K
Likes
64

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Matthijs/cmu-arctic-xvectors")