README.md

---
title: Malayalam Text → AI Voice (Free)
emoji: 🗣️
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 6.1.0
app_file: app.py
runtime: python
pinned: false
license: mit
models:
- ai4bharat/vits_rasa_13
---

# മലയാളം Text → AI Voice (Free)

A free web app (Hugging Face Space, Gradio) that converts **Malayalam** text to speech using the **AI4Bharat VITS** model.

## Features
- **Multiple voice styles** (ALEXA, NEWS, BOOK, etc.)
- **Prosody controls**: Speaking rate & pitch (approximate via resampling)
- **Batch paragraphs**: Split text by blank line → one file per paragraph × style
- **WAV + MP3** output (MP3 requires `ffmpeg`)

## Deploy / Run
1. Ensure the files below are present in the repository:
   - `app.py`
   - `requirements.txt`
   - `packages.txt` *(contains `ffmpeg` for MP3)*
   - `LICENSE`
2. Accept access to the gated model **ai4bharat/vits_rasa_13** on its model page (click “Access repository / Agree”).
3. If you still get permission errors, add a read token as a Space secret:
   - **Settings → Variables and secrets → New secret**
   - Name: `HF_TOKEN` | Value: your Hugging Face read token
4. Restart the Space.

## Notes
- Prosody controls are approximate (client-side resampling). For true SSML prosody, consider Azure AI Speech Malayalam neural voice (ml-IN-SobhanaNeural).
``