ggerganov/llama.cpp released b12452 with Fish Audio S2 support
Llama.cpp release b12452 introduces support for the Fish Audio S2 model, including built-in voice cloning capabilities. The update incorporates these new features into the core framework using an updated batch API. This addition expands the project's capabilities beyond standard text-based LLMs into advanced open-source text-to-speech generation. It allows developers running local AI workloads to leverage expressive voice synthesis and cloning directly within the ecosystem. The update, referenced under pull request #32330, utilizes the new batch API to clean up and integrate fish-audio-s2 seamlessly. It also packages precompiled binaries across a variety of target operating systems and hardware configurations.
## BACKGROUND
Llama.cpp is an open-source software library that performs inference on various large language models using the GGML tensor library. Fish Audio S2 is an expressive open-source text-to-speech model featuring multi-speaker support and fine-grained voice control.