Dama12/vits-tts-moore

VITS Mooré (Mooré Text-to-Speech)

This is a Text-to-Speech (TTS) model for the Mooré (Moore) language, a Gur language spoken in Burkina Faso by over 8 million people. The model is based on the VITS (Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech) architecture and trained using the Coqui TTS framework.

Model Details

• Developed by: Soumana Dama — Full stack developper & AI engineer-Founder & Lead AI Engineer at Scoinvestigator AI
• Model type: VITS (Text-to-Speech)
• Language(s) (NLP): Mooré (ISO 639-3: mos)
• Framework: Coqui TTS / PyTorch
• Parameters: ~28.6 Million (Generator) / ~74 Million (Total with Discriminator)
• License: Apache 2.0
• Developer Contacts:
  • LinkedIn: Soumana Dama on LinkedIn
  • Scoinvestigator AI: scoinvestigator.com
  • GitHub: Damasoumana1
  • Portfolio: soumanadama.netlify.app
  • Email: [email protected]

---

Uses

Direct Use

This model is designed to synthesize spoken Mooré audio from written Mooré text. It can be integrated into educational tools, voice assistants, accessibility software for the visually impaired, and content creation platforms.

Out-of-Scope Use

This model is trained on a specific Mooré dialect and voice. It is not intended for voice cloning without consent or generating deepfakes.

---

Training Details

Training Data

The model was trained on a custom, carefully curated Mooré speech corpus consisting of high-quality audio and their corresponding text transcriptions.

Training Procedure & Parameters

• Architecture: VITS (End-to-End TTS)
• Framework: Coqui TTS
• Hardware: Google Colab (Tesla T4 / A100 GPU)
• Training Steps: > 22,000 steps (~9,400+ epochs)
• Sample Rate: 16,000 Hz
• Metrics (at step 22k):
  • loss_gen (Generator Loss): ~35.8
  • loss_disc (Discriminator Loss): ~2.4

---

How to Get Started with the Model

To use this model, you will need to install the TTS library from Coqui.

Installation

pip install TTS

Inference in Python

Download the checkpoint_*.pth and config.json files from this repository and use the following Python code to generate audio:

import torch
from TTS.api import TTS

# Paths to your downloaded files
model_path = "checkpoint_22000.pth" # Replace with the latest checkpoint name
config_path = "config.json"

# Initialize TTS
# The model will automatically use GPU if available
tts = TTS(model_path=model_path, config_path=config_path, progress_bar=False)

# Text to synthesize
text = "A wilga a nin-buiidã a tʋʋm-kãsemsã."

# Generate audio and save to a file
tts.tts_to_file(text=text, file_path="output_moore.wav")

print("Audio generated successfully!")

Note on Special Characters

Ensure that the input text matches the orthography and character set used during training. Special characters not present in the vocabulary might be discarded by the phonemizer/tokenizer.

---

⚖️ License & Citation

License

This model is built upon the VITS architecture. The model weights are released under the CC BY-NC 4.0 License (free for academic and research purposes). The dataset, methodology, and this specific implementation are intended strictly for non-commercial research purposes. If you intend to use this model for commercial applications, please contact the author.

Citation

If you use this model or the Audio Mooré Project concept in your research, please cite it as follows:

@misc{soumanadama2026moore_tts,
  author = {Soumana Dama},
  title = {VITS Mooré: Text-to-Speech Model for the Mooré Language- Multilingual AI Assistant for Education in Burkina Faso},
  year = {2026},
  howpublished = {Hugging Face Model}
  note = {Applied Artificial Intelligence in the African Context}
}