The AI composer has two modes: Agent and Video Gen. Use Agent for a broader editing goal and changes to the project. Use Video Gen when you know the clip you want and prefer to set the model and inputs yourself.
Where do I switch modes?
Find the mode control at the bottom-left of the AI composer. Select the current mode name, then choose Agent or Video Gen. Switching changes the composer, not the active timeline. Recheck draft inputs after switching.
When should I use Agent?
Use Agent to assemble or restructure an edit, add generated elements in context, apply a Skill or Design Style, refer to exact project content, or combine several operations.
Write in the message field, attach files with +, or use the selection button to add precise @ references from the timeline, assets, Transcript, Viewer, or timeline ruler. You can also attach a Skill, style, voice, or video model from the icons below the field, then select Send.
Generation may show a card with Allow, Allow all … in this project, or Deny. Change an “allow all” choice under Agent Settings → Generation Auto-Allow.
When should I use Video Gen?
Use Video Gen for one direct video-generation job.
- Switch the composer to Video Gen.
- Select the model name in the bottom toolbar.
- Select Video settings and choose the input and output settings.
- Add the frame, references, or source video shown by that input mode.
- Write the prompt and select Send message.
The job appears in My Assets with progress. When complete, drag the result to a timeline. See Video Generation Models for current input, edit, and extension support.
What if generation takes a long time or fails?
Check its card in My Assets. Do not repeatedly submit the same request while the original is active. If it fails, review the visible message, correct rejected inputs, and retry or ask Agent to regenerate.