edwko/OuteTTS

Provides unified Python and JavaScript interfaces for running Llama-based OuteTTS speech synthesis models across multiple backends (llama.cpp, Hugging Face Transformers, ExLlamaV2, vLLM) with automatic sampling configuration and speaker profile management. Supports batched inference for real-time factor optimization on GPUs, with built-in tools for creating custom speaker profiles from audio samples and multilingual generation while preserving speaker characteristics.

1,429 stars. No commits in the last 6 months. Available on PyPI.