Tgiangvoice

🚀 Spark TTS Vietnamese

Spark-TTS is an advanced text-to-speech system leveraging large language models (LLMs) for highly accurate and natural-sounding voice synthesis. It's efficient, flexible, and powerful for both research and production.

🚀 Quick Start

Spark-TTS is an advanced text-to-speech system that harnesses the capabilities of large language models (LLMs) to achieve highly accurate and natural-sounding voice synthesis. It is engineered to be efficient, flexible, and powerful, making it suitable for both research and production purposes. This model is trained on the viVoice Vietnamese dataset.

📦 Installation

First, install the required packages:

pip install --upgrade transformers accelerate

💻 Usage Examples

Basic Usage

We have customized the code so you can perform inference using the Hugging Face Transformer library without installing anything else.

from transformers import AutoProcessor, AutoModel, AutoTokenizer
import soundfile as sf
import torch
import numpy as np

device = "cuda"
model_id = "DragonLineageAI/Vi-SparkTTS-0.5B"
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
model = AutoModel.from_pretrained(model_id, trust_remote_code=True).eval()
processor.model = model
 
prompt_audio_path = "path_to_audio_path" # CHANGE TO YOUR ACTUAL PATH
prompt_transcript = "text corresponding to prompt audio" # Optional
text_input = "xin chào mọi người chúng tôi là Nguyễn Công Tú Anh và Chu Văn An đến từ dragonlineageai"
 
inputs = processor(
    text=text_input.lower(),
    prompt_speech_path=prompt_audio_path,
    prompt_text=prompt_transcript,
    return_tensors="pt"
).to(device)
global_tokens_prompt = inputs.pop("global_token_ids_prompt", None)
 
with torch.no_grad():
    output_ids = model.generate(
        **inputs,
        max_new_tokens=3000,
        do_sample=True,
        temperature=0.8,
        top_k=50,
        top_p=0.95,
        eos_token_id=processor.tokenizer.eos_token_id,  
        pad_token_id=processor.tokenizer.pad_token_id  
    )
       
output_clone = processor.decode(
    generated_ids=output_ids,
    global_token_ids_prompt=global_tokens_prompt,
    input_ids_len=inputs["input_ids"].shape[-1]
)
 
sf.write("output_cloned.wav", output_clone["audio"], output_clone["sampling_rate"])

Advanced Usage

You can fine-tune this model with any dataset to improve quality or train on a new language. training code

📄 License

This project is licensed under the CC BY-NC-ND 4.0 license.