How Voicebox Works
Local + Online Setup Guide The FREE Open-Source Alternative to ElevenLabs
Download and install the Windows version from the official site or GitHub. The app itself is free and open source. https://voicebox.sh
Step 2: Launch the App
On first launch it will:
- Start the local backend
- Ask to download AI models
- Create a local data folder
The first model download may be several GB depending on which engine you choose.
https://docs.voicebox.sh/overview/installation?utm_source=chatgpt.com
Step 3: Download a Voice Model
Examples include:
- Qwen3-TTS
- Kokoro
- LuxTTS
- Chatterbox
- Hume TADA
These models are stored on your computer and reused offline.
Step 4: Clone Your Voice
Record around 10–30 seconds of clean speech.
Voicebox creates a reusable voice profile.
Rahul.wav
│
▼
Voice Embedding
│
▼
My Voice Profile
Step 5: Generate Speech
Type:
Hello everyone.
Welcome back to my channel.Choose your cloned voice.
Click Generate.
Voicebox creates an audio file using your voice.
Step 6: AI Agents Can Speak
If you're using an MCP-compatible AI agent (such as Claude Code with MCP), it can call Voicebox locally.
So instead of reading text, the AI responds with your chosen voice.