Some edits are easier
to point at than to say.

Conversational editing works well right up until you need to identify a specific thing in the frame. Then it falls apart, and it falls apart in a very particular way.

"Remove the object in the bottom right" (which object? There are four things in the bottom right. "The blue one") two of them are blue. "The one near the edge of the desk": you are now writing a paragraph to describe something you could point at in a quarter of a second.

This is not a flaw in language. It is that visual reference is what pointing is *for*, and prose is a bad substitute. Every editor knows the feeling of trying to give notes over the phone versus standing behind someone at the screen. Drawing on the frame is standing behind the screen.

Circle the region, then say what should happen:

• Remove it. A stray object, a logo you cannot clear, a person in the background, a boom mic in the top of frame.
• Blur or hide it. A whiteboard with someone else's roadmap on it, a screen with real customer data, a face that did not consent, a licence plate.
• Reframe around it. "Zoom in here" or "keep this in frame when you crop to vertical": the second is a genuinely hard problem to express any other way.
• Place something relative to it. "Put the caption above this" or "add an arrow pointing here."
• Emphasise it. Highlight, spotlight, or push in on the thing that matters at that moment.
• Fix it. "This is too dark" applied to a region rather than the whole picture.

The annotation tells Blob *what*; the sentence tells it *what to do*. Neither alone is enough, which is why the two work together.

Converting a horizontal video to vertical is where this pays off most, and it is a job almost everyone now has: a 16:9 recording that also needs to exist as a Reel, a Short and a TikTok.

Automatic reframing has to guess what matters. It usually tracks the face, which is right for a talking head and wrong for almost everything else. If the important thing is the product in your hands, the chart on the screen, or the second person in the shot, an automatic crop will cheerfully cut it out.

Circling the thing that must stay in frame removes the guesswork entirely. It is one gesture that replaces a conversation you would otherwise have to have shot by shot.

Blurring things is one of the most common real-world editing chores and one of the most annoying, because the thing to blur is usually incidental: you were not thinking about the whiteboard when you hit record.

Pointing at it is the natural interface for this. Circle the whiteboard, say "blur this," and it is handled for the shot rather than for one frame. The same goes for a screen showing real data, a document on the desk, a face in the background of a café, a package with an address on it.

The alternative (describing the location of a rectangle in words, or masking it by hand frame by frame) is exactly the kind of work that makes people not bother, and not bothering is how customer data ends up in a marketing video.