OuteTTS

OuteTTS is a local text-to-speech library for developers adding generated speech to their applications. It uses reference audio to reproduce a speaker's voice, including their style, emotion and accent.

• Reusable voice profiles: Create a speaker profile from an audio recording, save it and reuse it for later speech generation.
• Cross-language speech: Generate speech in another language using the same voice, though the reference speaker's accent can carry over into the result.
• Backend choice: Run OuteTTS models through llama.cpp, Hugging Face Transformers or ExLlamaV2, with GGUF support through llama.cpp.
• Batch generation: Process speech requests in batches through the asynchronous llama.cpp server and ExLlamaV2 backends, where model support allows it.
• Audio export: Save generated speech as WAV files for use in other applications.

The Python library is open source under Apache 2.0 and runs on your own hardware. It supports CPU execution and GPU acceleration through NVIDIA CUDA, AMD ROCm/HIP, Vulkan and Metal on compatible Macs, including Apple Silicon. A JavaScript package uses Transformers.js, but supports a narrower set of OuteTTS models.

The included test voice is English. Custom reference recordings affect output quality, and profiles recorded in the intended language generally give better results. Generation works best with short clips; reference audio takes up part of the available context, reducing how much speech a single run can produce.