Auto B-Roll, Explained: Manual Review vs. Fully Automatic

A clip that's just a face talking for 45 seconds can work, but cutting in relevant B-roll — even briefly — usually keeps attention better than a static shot. NimbleAI can find and insert that B-roll automatically as part of processing, in one of three modes.

No B-Roll

Exactly what it sounds like — the clip stays as-is, no automated footage insertion. Good for clips where the speaker's face and reactions are the point, like a reaction video or a close, personal story.

Manual Review

NimbleAI scans the transcript for moments where B-roll would help — a mentioned place, object, or concept — and pulls a handful of stock footage candidates for each moment. Instead of inserting them automatically, processing pauses and shows you the candidates so you can pick the best match, swap in your own upload, or skip that moment entirely. This is the default, and it's the right choice when you want the final say on what appears on screen.

Fully Automatic

Same detection, but instead of pausing for review, it renders directly with the first matching candidate for each moment. Faster, with less control — better suited to a high-volume workflow where you're processing a lot of clips and would rather review the finished product than every individual B-roll choice.

How the matching works

The detection step reads what's being said, identifies moments that reference something visual, and searches stock footage for a matching clip. It's matching concepts in the transcript to footage — not a perfect science, which is part of why Manual Review exists as the default rather than assuming the first automated pick is always right.

Which mode to use

  • Use No B-Roll for personal, reaction-driven, or talking-head-only content.
  • Use Manual Review when you want visual variety but care about exactly what's shown.
  • Use Fully Automatic when you're processing clips in volume and optimizing for speed over per-clip control.

Try B-roll on your next clip

Pick a mode in your dashboard and see how it changes a talking-head segment.

Try NimbleAI free
Back to all posts