audio-cpp

/

audio.cpp-gguf

Like 266

Hi @cbajtgvtk Native multi-speaker support depends on the model's capabilities. If a model supports native multi-speaker generation, you can use it directly through audio.cpp using the original model's input syntax, as with VibeVoce and Drama3.

Some downstream applications provide multi-speaker functionality by splitting text into chunks and assigning different voices to different chunks. However, this is an application-layer feature, not a native model capability. This functionality can certainly be built using audio.cpp as the inference engine, but the orchestration should remain at the application layer rather than the core runtime.