Datasetpublic

bamselarsen/Indextts_expansion

IndexTTS2 German Training Data Pre-processed German speech data for fine-tuning IndexTTS2. Contents File Description audio/ 18,597 WAV files (24kHz) manifest.jsonl IndexTTS-compatible manifest german_bpe.model SentencePiece tokenizer german_bpe.vocab Vocabulary file Stats Language: German (DE) Duration: 50 hours Samples: 18,597 Source: Emilia-YODAS dataset Quality filter: DNSMOS ≥ 3.0 Sample rate: 24kHz Usage #… See the full description on the dataset page: https://huggingface.co/datasets/bamselarsen/Indextts_expansion.

sourceHugging Facecc-by-4.0updated 9mo agoView on Hugging Face

0likes22downloads

Dataset Card

IndexTTS2 German Training Data

Pre-processed German speech data for fine-tuning IndexTTS2.

File	Description
`audio/`	18,597 WAV files (24kHz)
`manifest.jsonl`	IndexTTS-compatible manifest
`german_bpe.model`	SentencePiece tokenizer
`german_bpe.vocab`	Vocabulary file

Stats

—Language: German (DE)
—Duration: 50 hours
—Samples: 18,597
—Source: Emilia-YODAS dataset
—Quality filter: DNSMOS ≥ 3.0
—Sample rate: 24kHz

Usage

bash

# Clone the dataset
git clone https://huggingface.co/datasets/bamselarsen/Indextts_expansion

# Or download specific files
from huggingface_hub import hf_hub_download
hf_hub_download(repo_id="bamselarsen/Indextts_expansion", filename="manifest.jsonl", repo_type="dataset")

Manifest Format

Each line in manifest.jsonl contains:

json

{"id": "DE_xxx", "text": "German text...", "audio": "audio/DE_xxx.wav", "speaker": "...", "language": "de", "duration": 10.5}

Next Steps for Fine-tuning

1.Clone the index-tts repo
2.Download base IndexTTS2 checkpoints (gpt.pth, wav2vec2bert_stats.pt)
3.Run preprocessing: tools/preprocess_data.py
4.Train: trainers/train_gpt_v2.py

License

Audio data sourced from Emilia-YODAS dataset (CC-BY-4.0).