Logo Lanfrica

yibeltal-gashaw/amharic-text-to-speach

Domaine:

natural language processing

Type de record:

software
Créateur:
yib
Hôte:
# Amharic Text-to-Speech (TTS) Make written Amharic sound natural and clear. This project provides an end-to-end pipeline to convert Amharic text into speech, with a backend API and a web UI. ## Highlights - Clean, modern web interface for input and playback. - API-first design for easy integration into other apps. - Focused on Amharic language support and pronunciation. ## Repo Structure - `api/` Backend services and TTS processing. - `web/` Frontend application. ## Local Development 1. Start the API. 2. Start the web app. 3. Open the web UI and try a short Amharic phrase. ## Run Commands ```powershell # 1) install backend dependencies pip install fastapi uvicorn torch transformers pydantic # optional: better Amharic romanization support pip install uroman # 2) run API (localhost) uvicorn speech:app --host 0.0.0.0 --port 8000 --reload # 3) run web (tts_test.html) ``` ## Goals - Produce intelligible, natural-sounding Amharic speech. - Keep latency low for interactive use. - Provide a simple API for broader integrations. ## Status Active development. Feedback and contributions are welcome.