Build a local voice workflow you can trust.
Learn how to install LLMVoice, register reference voices, generate long-form speech, configure the pipeline, and understand the model license boundary.
Installation
Prepare Python, FFmpeg, PyTorch, and your compute device.
Quick Start
Generate your first local voice-cloned MP3.
Commands
The complete CLI command and flag reference.
Voices
Prepare, register, inspect, and manage reference voices.
Configuration
Set reliable defaults for each generation run.
Troubleshooting
Diagnose common environment and generation issues.
Licensing
Understand the MIT source and CPML model boundary.
What is LLMVoice?
LLMVoice is a local-first command-line tool for turning text transcripts into voice-cloned MP3 audio. It wraps voice management, XTTS-v2 generation, long-form chunking, audio merging, and diagnostics in one focused CLI.
Before you start
You will need Python, FFmpeg, PyTorch, and enough local compute for XTTS-v2. A CUDA-capable GPU is the fastest path, but CPU generation is supported.