|
|
--- |
|
|
title: Swahili Voice Assistant |
|
|
emoji: 🤖🇹🇿 |
|
|
colorFrom: green |
|
|
colorTo: blue |
|
|
sdk: gradio |
|
|
sdk_version: "4.31.0" |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
--- |
|
|
|
|
|
# 🤖 Msaidizi wa Sauti wa Kiswahili (Swahili Voice Assistant) |
|
|
|
|
|
This is a multi-modal Swahili assistant built with Gradio and Hugging Face Transformers. |
|
|
|
|
|
**Features:** |
|
|
- **Speech-to-Speech:** Talk to the assistant and get a spoken response. |
|
|
- **Text-to-Text:** Chat with the assistant using text. |
|
|
- **Tools:** Includes standalone tools for speech transcription and text-to-speech synthesis. |
|
|
|
|
|
**Models Used:** |
|
|
- **ASR (STT):** `EYEDOL/SALAMA_C3` (You must replace this placeholder in `app.py`) |
|
|
- **LLM:** `google/gemma-2-9b-it` |
|
|
- **TTS:** A custom ONNX model based on `facebook/mms-tts-swh`. |
|
|
|
|
|
**NOTE:** You must upload your `swahili_tts.onnx` file to this repository for the TTS functionality to work. |
