local models
01
transcription with Whisper
in the models tab you download a Whisper model (CoreML). after that local transcription runs fully on-device — no audio leaves the mac.
02
rewriting with llama.cpp
for local rewriting and embeddings rede bundles a llama.cpp server that only listens on 127.0.0.1. you pick the GGUF model in the models tab too.
03
all in one place
the “local models” window manages transcription, rewriting and embedding together: load, reload, delete — identical per model type.