Download

Settings

Settings: Local AI & Offline Models

Run transcription and caption generation entirely on your Mac with WhisperKit and MLX.

Run everything on-device: no internet, no per-minute costs, your files never leave your Mac.

WhisperKit

Download transcription models in Settings → Transcription. Larger models are more accurate; turbo is faster.

MLX

MLX is in-process local LLM caption generation (Qwen, Gemma, Llama, Ministral). No separate install needed.

RAM

MagiaCap recommends models based on your Mac's memory. 16GB comfortably runs large models.

Local models are private by default. Pair WhisperKit with an MLX model and you never touch the network.

In-app: open MagiaCap → Help → MagiaCap Help. All docs