Skip to main content
Not everything has to be generated. GoFrame Editor Agent can search stock libraries for real footage and photos, download and trim a YouTube video you have the rights to, ingest a folder of files, collect logos and screenshots from a product page, and look inside your footage to tell you what’s there. There’s no stock browser in the panels; you do all of this from the chat.

​
Before you start

  • A clear brief. Subject, action, mood and shot type: “barista pouring latte, close-up, moody” finds better clips than “coffee”.
  • Your project’s aspect ratio. Ask for vertical footage for a vertical project. Cropping landscape into 9:16 loses most of the frame.
  • Rights. Stock results come from libraries that license their media for commercial use, but you’re responsible for checking the license terms of anything you publish. Only download web videos you own or have permission to use.
  • Credits. Searching and downloading stock media doesn’t run a paid media model; the turn is billed as normal agent usage. Visual analysis uses an AI vision model; the agent announces any spending before it happens.

​
Find stock footage and photos

Start with /media-downloader or just ask.
Sources the agent can search: The agent picks a set with a consistent look (grade, era, lens feel) and tells you what it chose. Ask “show me more options” or “warmer, less corporate” to refine.

​
Where the result lands

Downloaded files are added to the Media pool like imports. Ask the agent to place them (“cut the three clips under the second paragraph of the voiceover”) or drag them to the timeline yourself. Stock clips aren’t marked Created by AI agent; the chat message and the asset’s Imported date are your record of where they came from.

​
Ingest a batch of files

/library-intake imports and tags a folder or list of files in one go, so the agent knows what’s in your library before it starts an edit.
For day-to-day imports use the Import button or drag files into the Media pool. See Import media.

​
Pull assets from a web page

Give the agent a URL and it can open the page, take screenshots, and extract the logo, product images, colors and copy. This is how the Create promo from URL suggestion and the Product promo Home template work: the wf-promo workflow builds a brand kit from the product page, proposes two or three promo directions, and generates the video.

​
See what’s in your footage

Before it makes editing decisions, the agent can look at your video with /visual: extract frames to view, analyze a segment in detail, find a moment (“where does she smile?”), detect shot boundaries, or check whether a generated clip really moves or is just a slideshow.
Results come back as text and frame thumbnails in the chat. Timecodes refer to the source file, and the agent uses them to add markers, cut clips or align B-roll.

​
Tips

  • Ask for a specific number and length (“3 clips, 10–20 s each”) so the agent doesn’t have to guess.
  • Broaden a search by dropping words, not by switching language.
  • Don’t mix stock clips with visibly different color temperatures in one sequence unless you plan a grading pass. Ask the agent to “match the look of these three clips”.
  • Stock is for real-world B-roll. For an impossible shot or a consistent invented character, generate video instead.