Dataset · ylacombe
Multilingual Audiobook TTS Audio Dataset
ylacombe/cml-tts
Massive audiobook speech corpus in six European languages for text-to-speech tasks.
What's inside
Contains pairs of audio recordings and texts numbering between 1 and 10 million segments. Recordings are sourced from LibriVox audiobooks read by volunteers from Project Gutenberg public domain books. The dataset covers Dutch, German, French, Italian, Polish, and Portuguese.
What you can do with it
- Training and fine-tuning multilingual text-to-speech (TTS) models
- Training automatic speech recognition (ASR) models on European languages
- Voice cloning and natural reading intonation modeling
- Tasks
- Text → speech, Text → audio
- Languages
- Dutch, French, German, Italian, Polish, Portuguese, Spanish
- Size
- 1M–10M rows
- Downloads this month
- 56.1K
- Likes
- 50
Models trained on it
- alien79/F5-TTS-italianText → speech 1.4K
- gian2007/F5-TTS-italianText → speech 129
- hr16/multilingual_spinEmbeddings 124
- tts-hub/yourtts_cmlText → speech 79
- GatinhoEducado/speechT5_tts-finetuned-cml-ttsText → speech 48
- TheMiniatureGamer/OmniVoice-Polish-FinetuneText → speech 25
- SamuelM0422/speecht5_finetuned_cml-tts_ptText → audio 20
- Gregniuki/F5-tts_English_German_Polish 0
How to load it
from datasets import load_dataset
ds = load_dataset("ylacombe/cml-tts")