---
title: VoiceVerse AI
emoji: 🎙️
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: "5.23.1"
python_version: "3.10"
app_file: app.py
pinned: false
---

# 🎙️ VoiceVerse AI — Document to Audio

Transform uploaded documents into engaging, emotionally expressive podcast-style audio narrations.

## Pipeline

```
PDF/TXT → Text Extraction → RAG (chunk + embed + retrieve) → Script Generation (Mistral-7B) → TTS (Qwen3-TTS / Edge-TTS) → Audio Playback
```

## Models Used

| Component | Model | How |
|-----------|-------|-----|
| Embeddings | `all-MiniLM-L6-v2` | Local (CPU) |
| Script Gen | `Mistral-7B-Instruct-v0.3` | HF Inference API |
| TTS (primary) | `Qwen3-TTS` | HF Inference API |
| TTS (fallback) | `Edge-TTS (AriaNeural)` | Local (CPU) |

## Setup

```bash
pip install -r requirements.txt
export HF_TOKEN="your_huggingface_token_here"
python app.py
```

## Deployment on HF Spaces

1. Create a new Space (Gradio SDK)
2. Upload all project files
3. Set `HF_TOKEN` as a Space Secret
4. The app will auto-launch on port 7860

## Project Structure

```
app.py           # Gradio UI entry point
rag.py           # Document ingestion, chunking, embedding, retrieval
script_gen.py    # LLM script generation (Mistral-7B-Instruct)
tts.py           # Text-to-speech (Qwen3-TTS + Edge-TTS fallback)
utils.py         # Helpers (temp files, validation, error formatting)
requirements.txt # Python dependencies
packages.txt     # System packages (ffmpeg)
```