Hindi VITS TTS (V1 Baseline)

This is an end-to-end Text-to-Speech model for Hindi based on the VITS architecture.

Model Details

  • Architecture: VITS (Variational Inference with adversarial learning)
  • Language: Hindi (Devanagari)
  • Status: Research Baseline (V1)

Training Data

Trained on the IIT Madras IndicTTS Database, specifically the Hindi monolingual subset:

  • Speakers: 2 (1 Male, 1 Female)
  • Sampling Rate: 22,050 Hz
  • Content: Standard Hindi utterances with high-quality studio recordings.

Features

  • Handles Devanagari Unicode normalization.
  • Number-to-word conversion for Hindi digits.
  • Trained on Kaggle GPU environments.
Downloads last month
1
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support