Skip to Content
DocumentationRush Agent

Rush Agent

The Rush Agent is an AI assistant built into the editor. Instead of manually searching for footage, uploading files, or clicking through settings, you tell the agent what you want in plain language and it makes the change for you. Swap a clip, pull a fresh video from the web, regenerate a stretch of voiceover, strip the effects out of a project you ask, the agent does it.

The Rush Agent is in early beta. It is powerful for small, targeted edits and improves with every update, but it is not yet built for large, sweeping changes. This page covers what it does well today, how to phrase your requests, and where its current limits are.

💡

If the Rush Agent ever gets stuck or stops responding mid-task, start a new chat and try again.


What the Rush Agent can do

Replace footage

The most common use. Point the agent at a clip you don’t like and describe what should go there instead:

“Replace this image with a video of the MC-25 flying through the air.”

The agent searches for matching footage on the web, pulls it, and drops it straight onto your timeline. No manual uploading, no searching stock libraries yourself.

By default the agent sources existing footage from the web; it does not generate anything. If the first result misses, ask again with a correction (“this isn’t quite right, find a video of a real Soviet-era fighter jet”). See Broaden your search below when a specific object is hard to find.

Generate new footage

The agent only creates footage when you explicitly ask it to generate. A plain “replace this” always sources real footage; say “generate a new image/video of…” when you want something made rather than found.

Regenerate voiceover

One of the most requested features. If a stretch of narration has a glitch or a bug, ask the agent to redo just that segment:

“There’s an issue with the voiceover between 0:00 and 0:30 at the start of this video. Please regenerate it for me.”

The agent regenerates that voice segment in place and fixes the issue, usually in seconds. You no longer need to contact support or re-export the whole video for a single voiceover bug, point the agent at the timestamps and let it rebuild that piece.

Adjust motion graphics

If you don’t like a specific motion graphic, ask the agent to edit or change it rather than hunting through settings yourself.

Change project settings

The agent can act on the project as a whole, beyond individual clips. For example:

“Remove all the effects in this video.”

It applies the change across the project without you opening each menu.


How to phrase requests

The Rush Agent works best with specific, bounded instructions.

  • Point to a clip or a timestamp range. “Replace the two clips between 0:00 and 0:06 with their still-image equivalents. Find images that match the same type of footage.”
  • You can target multiple clips at once. Asking it to replace several specific clips in one request is fine.
  • Keep the scope small. Around 30 seconds of a sequence is a comfortable working range for replacing footage, adjusting pacing, or changing effects.

Current limits (early beta)

The Rush Agent is still in early beta, so keep tasks modest for now.

  • Don’t hand it long sections. Editing an entire one- to three-minute stretch in a single request will often produce mistakes. Break big changes into smaller ones.
  • Long-horizon edits aren’t ready yet. “Re-edit the first three minutes of this video” is beyond what the agent can reliably do today. Longer-horizon regeneration is coming.
  • Very specific objects can be hard to source. The agent does better on general items and images than on rare, highly specific subjects.

If the agent can’t find footage of a very specific object (say, the MC-25 fighter jet), widen the request. Ask for the broader category instead, “a Soviet-era fighter jet” rather than the exact model. A broader query gives the agent far more footage to choose from and usually lands a good match faster.


What’s coming

The Rush Agent will become more agentic over time: re-editing longer parts of your video, adjusting pacing across whole sequences, and handling bigger, multi-step tasks. For now it shines at clip replacement, voiceover regeneration, and small adjustments, and the team would love your feedback as it develops.

Last updated on