---
language: lg
license: apache-2.0
tags: [text-to-speech, tts, nemo, luganda, african-languages]
datasets: [Sunbird/salt]
---

# Luganda TTS v3

Text-to-Speech system for Luganda using NVIDIA NeMo.

## Models
| Model | Description | Size |
|-------|-------------|------|
| `luganda_fastpitch.nemo` | FastPitch spectrogram generator | 187 MB |
| `luganda_hifigan.nemo` | HiFi-GAN neural vocoder | 339 MB |

## Training
- **Dataset**: Sunbird/salt (~2,380 samples, 2.69 hours)
- **FastPitch**: 20,000 steps
- **HiFi-GAN**: 20,000 steps
- **Sample Rate**: 22,050 Hz

## Usage
```python
from nemo.collections.tts.models import FastPitchModel, HifiGanModel

fastpitch = FastPitchModel.restore_from("luganda_fastpitch.nemo")
hifigan = HifiGanModel.restore_from("luganda_hifigan.nemo")

text = "Oli otya?"
spec = fastpitch.generate_spectrogram(tokens=fastpitch.parse(text))
audio = hifigan.convert_spectrogram_to_audio(spec=spec)
```