🎙️ RVC Text-to-Speech Studio

Convert any text into a custom voice using edge-tts and RVC (Retrieval-based Voice Conversion).
Built with Streamlit for a sleek, interactive, and modern web interface.

---

✨ Features

• 🗣️ 13+ Base Voices — Choose from Turkish, English, German, French, Spanish, Japanese, Korean, and Chinese neural voices via edge-tts.
• 🎤 RVC Voice Conversion — Transform the base speech into any custom voice with a trained .pth model.
• 🎼 Multiple F0 Methods — rmvpe (best quality), harvest, crepe, pm (fastest).
• 🎵 Pitch Control — Adjust pitch ±24 semitones to match the target model's range.
• ⚙️ Fine-Tuning — Index rate, RMS mix rate, consonant protection sliders.
• 🖥️ GPU & CPU Support — Automatically uses CUDA if available, falls back to CPU.
• ⬇️ Download — Listen in-browser and download the final WAV file.

---

🚀 Quick Start

1. Clone the Repository

git clone https://github.com/ErenBalkis/rvc-tts-studio.git
cd rvc-tts-studio

2. Create a Virtual Environment (Python 3.10 Recommended)

python -m venv venv

# Windows
venv\Scripts\activate

# macOS / Linux
source venv/bin/activate

3. Install PyTorch

Warning

You must install the correct PyTorch build before installing the other dependencies.

Hardware	Command
CPU only	pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
CUDA 11.8	pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
CUDA 12.1	pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

4. Install Remaining Dependencies

pip install -r requirements.txt

5. Add Your RVC Models

Create a subfolder for each voice inside the models/ directory. Each subfolder should contain a .pth file and optionally a .index file:

models/
├── my_voice/
│   ├── my_voice.pth
│   └── my_voice.index   ← optional, improves quality
├── another_voice/
│   └── another_voice.pth
└── .gitkeep

Tip

Where to find RVC models? Search for pre-trained RVC models on Hugging Face or weights.gg.

6. Run the App

streamlit run app.py

The app will open at http://localhost:8501 🎉

---

🎯 Usage Guide

1. Select a model from the sidebar dropdown (click 🔄 Refresh Models after adding new files).
2. Type or paste your text in the main area.
3. Choose a base voice — the edge-tts speaker used before RVC conversion.
4. Adjust the pitch slider to match your target model's vocal range (positive = higher, negative = lower).
5. Fine-tune quality settings in the sidebar (F0 method, index rate, RMS mix, protect).
6. Click 🚀 Generate Voice and wait for the result.
7. Listen in-browser or ⬇️ Download the final WAV.

⚙️ Quality Settings

Setting	Description	Recommended
F0 Method	Pitch extraction algorithm	rmvpe for best quality
Index Rate	How much target voice character to apply (0.0–1.0)	0.75
RMS Mix Rate	Volume envelope mixing (0.0–1.0)	0.25
Protect	Consonant protection against artifacts (0.0–0.5)	0.33
Pitch Shift	Semitone adjustment (-24 to +24)	0 (adjust per model)

---

🛠️ Troubleshooting

Problem	Solution
numpy version conflict	Make sure numpy < 2.0 is installed: pip install "numpy<2.0"
torch not found	Install PyTorch manually using the table in Step 3 above
No models in dropdown	Place .pth files inside subfolders in models/ and click Refresh Models
CUDA out of memory	Use a smaller model or switch to CPU (torch CPU build)
rvc-python import error	Run pip install rvc-python>=0.1.5

---

📁 Project Structure

rvc-tts-studio/
├── models/            # Your RVC model subfolders (.pth & .index)
│   └── .gitkeep       
├── temp/              # Temporary audio files (auto-managed)
│   └── .gitkeep
├── app.py             # Streamlit UI & application logic
├── rvc_module.py      # RVC inference & model discovery module
├── requirements.txt   # Python dependencies
├── .gitignore         
└── README.md          

🛡️ Tech Stack

• Streamlit — Interactive web UI
• edge-tts — Microsoft Edge's online TTS engine
• rvc-python — RVC inference wrapper
• PyTorch — Deep learning backend
• librosa — Audio processing