Hindi TTS (Text-to-Speech, 24kHz)
Overview
Hindi TTS is a high-quality Text-to-Speech model developed using the F5 TTS architecture. Built by FuturixAI and Quantum Works, this model enables natural-sounding Hindi speech synthesis and is distributed under the MIT license. It is intended for both research and commercial applications.
Key Features
- Language: Hindi
- Sampling Rate: 24 kHz
Training Data
The model was trained on the IndicVoices-R_Hindi dataset provided by IIT Madras.
Usage Instructions
Prerequisites
Ensure you have installed the necessary dependencies for the f5-tts_infer-cli. Refer to the GitHub repository for installation instructions:
https://github.com/rumourscape/F5-TTS
Example Usage
f5-tts_infer-cli \
--model "Futurix-AI/Hindi-TTS" \
--ref_audio "ref_audio.wav" \
--ref_text "यह संदर्भ ऑडियो का सामग्री, उपशीर्षक या लिप्यंतरण है।" \
--gen_text "यह एक उदाहरण है जो मॉडल से बोलने के लिए उत्पन्न किया गया है।"
Parameters:
--model: Replace "hindi_tts_checkpoint.pth" with the actual checkpoint file name.--ref_audio: Path to the reference audio file (e.g., "ref_audio.wav").--ref_text: Hindi text corresponding to the reference audio.--gen_text: Hindi text for the TTS model to generate speech.
license: mit
Inference API (serverless) does not yet support safetensors models for this pipeline type.
Model tree for Futurix-AI/Hindi-TTS
Base model
SWivid/F5-TTS