Use LM Studio (local AI)

LM Studio is a desktop app for running AI models on your own computer. Evaloon Editor can use it as the brain of the AI Editor, with no API key.

Set it up

Install LM Studio from lmstudio.ai and download a model inside it. Choose one with the hammer badge, which means it was trained to use tools. Without that, it cannot edit your timeline.
In LM Studio, start its local server.
In Evaloon Editor, open Settings, choose the AI Editor tab, and set Provider to LM Studio (local).
Leave Base URL as http://localhost:1234 (LM Studio’s usual address). Choose your model from the Model list.
Read the status line under the model. When it says LM Studio is ready, open the AI Editor panel and try a request.

Why “8K” keeps showing up

LM Studio decides how much text a model can hold (its context size) when the model loads, not for each request. Evaloon Editor needs 8K of context. So the app loads your model at 8K and checks this before each turn.

If the model is already loaded at a smaller size, the app asks you before reloading it, and the status line offers Reload at 8K. It never reloads on its own, in case another program is using that model.

What the status line can tell you

Message	What to do
LM Studio isn’t reachable at …	Open LM Studio and start its local server, then Check again.
… isn’t downloaded	Download it in LM Studio, or choose another model.
… isn’t trained for tools	Choose a model with the hammer badge.
… holds at most a … context	That model is too small for the editor. Choose another.
… isn’t loaded yet	Nothing to do. The app loads it with an 8K context when you send a request.
… is loaded twice	Unload one copy in LM Studio.
… is loaded with a … context, and the app needs 8K	Select Reload at 8K.

Tips

Only one local provider can keep a model in video memory on a small PC. If you also use Ollama, unload its model first.
Local models are good at templates, captions and removing silences. For harder edits, see Cloud keys.