EMTIAZZ/bangladeshi-bangla-tts-vits

Bangladeshi Bangla Text-to-Speech (VITS)

This is a fine-tuned VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) model for Bangladeshi Bangla text-to-speech synthesis.

Model Description

• Language: Bangla (Bengali) - Bangladeshi variant
• Architecture: VITS (Variational Inference Text-to-Speech)
• Task: Text-to-Speech (TTS)
• Training: Fine-tuned on Bangladeshi Bangla speech data
• Voice: Female speaker
• Sample Rate: 22050 Hz

Features

• High-quality Bangla speech synthesis
• Support for Bangladeshi Bangla pronunciation and intonation
• End-to-end neural text-to-speech
• Real-time inference capability
• Proper handling of Bangla numerals and punctuation

Usage

Using with TTS Library

import torch
import soundfile as sf
from TTS.api import TTS

# Load the model
tts = TTS(model_path="path/to/pytorch_model.pth", 
          config_path="path/to/config.json")

# Generate speech
text = "আমি বাংলা টেক্সট টু স্পিচ ব্যবহার করছি।"
audio = tts.tts(text)

# Save audio
sf.write("output.wav", audio, 22050)

Using with Hugging Face Hub

from huggingface_hub import snapshot_download
import torch
import soundfile as sf

# Download model
model_path = snapshot_download(repo_id="EMTIAZZ/bangladeshi-bangla-tts-vits")

# Load and use the model
# (Add your custom loading code here)

Model Performance

• Training Epochs: 100
• Model Size: ~997MB
• Inference Speed: Real-time capable
• Audio Quality: High fidelity speech synthesis

Training Data

This model was fine-tuned on Bangladeshi Bangla speech data to capture the specific pronunciation, intonation, and linguistic characteristics of Bangladeshi Bengali.

Limitations

• Optimized for Bangladeshi Bangla variant
• Single female voice
• May not handle out-of-vocabulary words perfectly
• Performance may vary with very long texts

Technical Details

• Framework: PyTorch
• Architecture: VITS
• Sampling Rate: 22050 Hz
• Quantization: Model is quantized for efficient inference

Citation

If you use this model in your research, please cite:

@misc{bangladeshi-bangla-tts-vits,
  title={Bangladeshi Bangla Text-to-Speech using VITS},
  author={EMTIAZZ},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/EMTIAZZ/bangladeshi-bangla-tts-vits}
}

License

This model is released under the Apache 2.0 License.

Contact

For questions or issues, please open an issue in the repository or contact the author.