Skip to content
Evaloon
Download

Edit with the AI chat

The AI Editor is a chat panel. You say what you want (“trim the first clip to 5 seconds, add a fade-in, and put a title card at the start”) and the AI makes those changes on your timeline.

It can only edit once a model is connected. You pick one provider and the app remembers it.

You want Choose Guide
Free, private, on your own PC Ollama (local) or LM Studio (local) Ollama, LM Studio
The strongest edits Anthropic (Claude), OpenAI, Gemini or OpenRouter, with your own key Cloud keys
To use Claude you already have Claude (your Claude Code install) Cloud keys
To drive the editor from another tool The MCP server Connect Claude Code

Local models handle templates, captions and silence removal well. Choosing what to cut by content, and writing custom animation, works best with a stronger model such as Claude, OpenAI or Gemini.

  1. Open a project, then open the AI Editor panel.

  2. If no model is connected, the panel shows a Connect a model button. Select it. (You can also open Settings and choose the AI Editor tab.)

  3. Pick a Provider and fill in what it asks for. The guides above cover each one.

  4. The panel header shows the model you chose. The AI settings button there lets you change it later.

Type in the box and press Send. Good first requests:

  • “Add a title that says ‘Welcome’ for the first 3 seconds”
  • “Trim 2 seconds off the end of the first clip”
  • “Add a fade-in to the opening clip”
  • “List everything currently on my timeline”

Be specific about what and where. “Make it better” is too vague. “Remove the silences and add captions” is clear.

  • Approval cards. Before it deletes something, costs money, cuts words, removes silences or adds captions, the chat asks first. Choose Run (or Approve) to go ahead, or Reject.
  • Auto-approve (shield icon) skips the questions. Turn it on only for a project and model you trust.
  • Dry-run (flask icon) shows what the AI plans to do without changing anything.
  • Undo last AI turn reverts everything the AI did in one reply as a single step.
  • Stop ends the current reply and rolls back any half-finished edits.
  • Token meter shows how much text has gone to and from the model in this chat. Cloud models charge by that amount.
  • New chat and Conversation history start over or bring back an earlier chat.

With Ollama or LM Studio, the chat runs on your own computer. With a cloud provider, your requests and the project details the AI needs go to that provider, using your key. Pictures of your timeline are only sent when the AI asks to look at a frame, and the chat records when that happens.