Skip to content

Local AI with Ollama or LM Studio

Zush runs local analysis through Ollama or LM Studio on Mac and Windows. You choose and run a compatible model in the local server, then connect Zush to it. Supported analysis stays on your computer while the local mode is selected.

  • the latest Zush version for Mac or Windows
  • macOS 15 or later, or Windows 10/11
  • enough disk space and memory for a compatible local model
  • Ollama or LM Studio on the same computer; the setup guides below cover installation
  • internet access for installation and model download; analysis can run offline afterward
  1. Open Zush Settings and select AI Modes.
  2. Choose Ollama or LM Studio.
  3. Follow the matching setup guide below. For Ollama, Zush provides buttons to download Ollama, start it, and download the first model. For LM Studio, install and load a vision-capable model in LM Studio.
  4. Confirm the selected model and use Test to check the connection.
  5. Open AI Rename and test a small representative batch.
OptionUse it when
OllamaYou want guided setup using Zush’s Download and Start buttons, or already use Ollama
LM StudioYou want a desktop model catalog and server controls

Local modes use the previews, extracted text, and metadata Zush can prepare. They do not send raw audio to the local model or automatically use cloud transcription. See Audio and video analysis before choosing a mode for recordings.

ProblemWhat to do
Zush cannot connectConfirm that Ollama or LM Studio is running and that the server address matches Zush Settings
The model is not listedLoad or pull a compatible model in Ollama or LM Studio, then refresh the model list in Zush
Analysis is slowTry a smaller compatible model or close other memory-intensive apps
A file type failsTest a supported file with a vision-capable model and check the relevant local-server logs