The ggml-org team has released llama.cpp b12426, adding support for Higgs v3 speech-to-text and VibeVoice-ASR in its media-handling code. Speech-to-text converts spoken audio into written words.
The release page also provides downloadable builds for macOS, iOS and several Linux systems. It does not specify which model versions or usage limits apply to the new audio support.
Why it matters
Developers using llama.cpp can try its new support for these speech models, though the release notes do not explain the setup requirements or supported model details.