OngeaLabs is a lightweight Text-to-Speech (TTS) studio for African languages built in Python + Streamlit. It supports single-text speech generation and batch line-by-line clip generation, offers multiple voice models per language (Meta MMS + community voices), includes tone controls (speed/pitch), saves WAV outputs to disk, and keeps an in-session audio library for quick playback and reuse.
Live app: https://ongealabs.streamlit.app/
- Single Text → Speech
- Generate one-off voice clips from a text prompt.
- Batch Clip Generator
- Paste multiple lines and generate clips line-by-line (ideal for voiceovers).
- Multi-Voice, Multi-Language
- Select from Meta MMS voices and community voice models per language.
- Tone Controls
- Adjust speed and pitch to match narration style.
- WAV Outputs + Session Library
- Saves generated audio as .wav files.
- Keeps an in-session library for playback and quick access.
- Local Fine-Tuning Launcher
- Can launch local fine-tuning using a Hugging Face dataset and the finetune-hf-vits training repo.
This app is hosted on Streamlit Community Cloud and may go to sleep when idle.
If prompted, click “Wake this app” and retry after it starts.
- Python
- Streamlit
- Meta MMS (multilingual speech models/voices)
- Hugging Face Datasets (for training data)
- finetune-hf-vits (training workflow)
Your repo may differ — adjust this section to match.