Insert B-roll

Overlay stock or imported footage at the exact moments your speaker mentions them. Search, import to bin, place — one prompt.

A talking head gets 2–3x more watchable when the visuals change while the person speaks. Ask the agent for B-roll and it finds footage, matches it to what's being said, and overlays it at the right moments — no manual searching, downloading, or nudging clips into place.

Prompts to paste

"Add B-roll where it makes sense — overlay stock footage of the concepts I'm talking about."

"Find city-at-night stock clips and overlay them while I talk about the launch, around 0:20 and 0:45."

"Use my unused imported clips as B-roll at natural moments."

"Insert product close-ups as B-roll — 2 seconds each, no more than 3 placements."

"Dress this up: B-roll on the abstract parts, but keep my face on screen for the intro and the CTA."

What the agent actually does

The full pipeline is two stages working together:

  1. Search and import — the agent searches the built-in Pexels stock library and imports results into your media bin without placing anything on the timeline.
  2. Place — it lays imported image and video assets onto overlay tracks, either at moments you specify (in seconds) or at transcript beats and main-track intervals when you don't. Give it a goal (e.g. "product", "customer", "dashboard") and it matches asset names against transcript beats so the right clip lands on the right sentence.

If you already imported footage yourself, the agent skips step 1 and uses your unused bin assets directly.

Defaults: up to 4 placements, 2.5 seconds per overlay. Both are adjustable — just say so in the prompt.

Where the footage comes from

  • Stock — the agent searches Pexels (free, commercially usable). Videos for motion, photos for static concepts. Needs a stock provider key on the server; without one, stock search degrades gracefully and the agent falls back to your imported assets.
  • Your imports — drag files into the Media panel first (Import media). The agent can't import local files by path — it only sees what's in the bin.

Descriptive filenames help: rename clips to "product-closeup.mp4" or "customer-smiling.jpg" so goal matching picks the right ones.

When to do it manually

Drag an asset from the Media panel onto an overlay track above your main clip, then trim to length — see Import media and Timeline editing.

Limits

  • B-roll goes on overlay tracks — it covers the main video but doesn't replace or re-cut it. Ask the agent to flatten the overlays if you want a sequential cut instead of picture-in-picture.
  • Asset-to-moment matching is heuristic (names + transcript beats), not visual understanding of the B-roll itself.
  • Placement at "natural moments" needs a transcript; without one the agent spaces overlays across the main track.
  • Overlays can cover captions — ask the agent to check, or nudge them on the timeline.

See also

PreviousCaption skinsNextAuto sound design
Community