Itech8234/hausa-tts

# Muryar Hausa — Hausa Text-to-Speech A Flask + htmx web app that converts Hausa text into speech using two backends: - **gTTS** — Google Translate's TTS API (`lang='ha'`). Fast, needs internet, no setup cost. - **MMS-TTS** (`facebook/mms-tts-hau`) — Meta's neural VITS model fine-tuned for Hausa. More natural-sounding, runs locally/offline once downloaded, but heavier (needs `torch`). ## Features - Single-sentence synthesis with a live progress bar - **Batch mode**: paste up to 25 lines, each becomes its own audio clip, with a real per-item progress bar streamed over Server-Sent Events, plus a "download all as .zip" button - Session-based history panel with replay/download - Per-IP rate limiting (in-memory, no Redis needed) - Automatic cleanup of generated audio files older than `AUDIO_MAX_AGE_MINUTES` - Dark theme with a subtle CSS-only Islamic geometric pattern background ## Project structure ``` hausa-tts/ ├── app.py # Flask routes, rate limiter, cleanup thread, batch/SSE ├── tts_engine.py # gTTS + HF MMS-TTS wrapper ├── requirements.txt # full deps (includes torch/transformers for MMS-TTS) ├── requirements-render.txt # lean deps for constrained hosts (gTTS only) ├── Procfile # gunicorn start command ├── render.yaml # Render blueprint (one-click deploy config) ├── templates/ │ ├── base.html │ ├── index.html │ ├── _result.html # htmx partial for single synthesis result │ ├── _batch_result.html # initial batch container (opens SSE connection) │ ├── _batch_progress.html # SSE-streamed progress fragment │ ├── _history.html │ └── _error.html └── static/ ├── css/style.css └── audio/ # generated mp3/wav/zip files land here ``` ## Local setup ```bash python -m venv venv source venv/bin/activate # Windows: venv\Scripts\activate pip install -r requirements.txt python app.py ``` Then open `localhost`. ### First run with MMS …