The ggml-org team has added support for the Zonos3-Mini text-to-speech model in llama.cpp release b12424. The update includes an audio-generation pipeline, voice cloning, text normalization and delivery parameters. It also supports storing a default voice in GGUF model metadata, a format used to package models. The release page lists downloadable builds for several platforms, including macOS, iOS and Linux.
Why it matters
People using llama.cpp can now try Zonos3-Mini speech generation in the project, though the release notes do not specify which voices or languages it supports.