What is LuviBot and what can it do?

Three modes

LuviBot opens in the Workspace and shows three buttons in its header:

• Model Advisor. Describe the job — "a short promo clip for social media", "realistic product photos" — and it recommends a primary model and alternatives, with a one-line reason for each and a button to select the model straight into the Workspace.
• Image Prompt Engineer. Give it a rough idea and it writes a complete image prompt in the dialect of the model you selected, including a negative prompt where the model supports one. Copy copies both; Use drops the prompt into the prompt box.
• Video Prompt Engineer. The same for video: camera motion, atmosphere and pacing written the way the selected video model expects.

When you open the image editor, LuviBot switches itself into an edit-assist mode and reads the picture and the mask you are working on.

How it behaves

• It starts closed. A small bubble invites you a few seconds after the Workspace opens; it disappears on its own.
• Replies come in the interface language you are using.
• Conversations are saved, so you can reopen an earlier chat from the history icon in its header or start a new one.
• Nothing it writes is sent to a model until you press Generate.

Tips

Tell it the format ("vertical, 9:16") and the mood; it turns that into parameters and wording. For models with a prompting guide, the Workspace's own prompt check will still run on the text it wrote.