Auto reframe with multiple speakers: why the resize crops the wrong head out
Quick Answer
Auto reframe with multiple speakers describes what happens when software converts widescreen video to vertical or square by tracking a subject and cropping around it. With one person centred it holds up. With two people in frame the software has to pick a subject, and it often picks the wrong one and cuts a head off.
What does auto reframe do with a single centred speaker?
It handles that case most reliably: one person, roughly in the middle, facing the camera. The software tracks that face, keeps it inside the new crop, and the resize survives the trip from widescreen down to vertical or square with nothing important lost.
Move the same speaker to one side of the frame, next to a slide or a product on a table, and the software still finds the one face it has to track. Whether it keeps the product in frame alongside them is a separate question, and often the answer is no. Planning for that starts before the edit, which is the case for thinking about framing decisions at the point of capture rather than after the fact.
What breaks when auto reframe meets multiple speakers?
With two people in the shot the software has to choose a subject, and nothing tells it which one carries the video. A video production team lead at an enterprise communications vendor put it plainly:
If, if there are like two speakers in, on the video, yeah. When it does the resize, the subject head was crop out
The same person raised the harder version, the one with no clean answer:
what if there is a, several people, in the shot, how frame, the correct one?
Tracking one subject is a technical problem software solves well. Choosing between two subjects is a judgment about who the audience came to see, and that judgment needs somebody who knows what the video is for.
| Shot type | What auto reframe does | What goes wrong | What a human still decides |
|---|---|---|---|
| Single speaker, centred | Tracks the one face and holds it in the new crop | Little; this is the case it handles most reliably | Whether anything outside the face, like a logo or a slide, has to stay visible |
| Single speaker, off centre | Follows the face even when it starts near the edge | Loses context around the speaker, such as a product or a second element | Whether that surrounding context matters more than a tight crop |
| Two speakers | Picks one face to track as the primary subject | The other speaker's head gets cropped out, sometimes mid sentence | Which speaker the audience needs to see, and when |
| A speaker who moves | Tries to keep tracking as the subject shifts position | The subject goes half out of frame, then back in, as they move | Whether to recentre by hand or plan the framing differently from the start |
Why did two separate teams hit the same auto reframe problem?
Because the ambiguity is in the footage rather than in any one product. A marketing lead at a legal technology company ran into the identical issue six months later, in a different industry, converting an interview clip into a vertical social format:
he's like he's half out of frame, right? And then he's back in and I'm kind of like we could just like move this over.
The same person asked for the obvious remedy:
Is there any way to center the video that gets cropped
Two accounts in two industries described the same failure in nearly the same words, six months apart. Aspect ratio turns out to be a brand decision as much as a technical one, and worth settling before anyone is troubleshooting individual clips: see the case for treating aspect ratio as a brand rule.
Who makes the final call when auto reframe cannot pick a subject?
Somebody has to decide which speaker holds the frame, and that decision belongs close to the footage rather than after the crop has already shipped. Take the interview clip from the legal technology account. It was shot for one format, needed to work in another, and had two people moving in and out of frame as they talked. Auto reframe made a reasonable guess and got it half right, and somebody on that team went back in and moved the crop by hand.
Two things make that correction cheaper. If the destination is paid social, the crop requirements are already spelled out by the platform, which is covered in the rundown of specs for paid media formats. And how much resolution you capture up front decides how much room you have to reposition later, which is the subject of what resolution to shoot at for downstream cropping.
FAQ
What is auto reframe in video editing?
Auto reframe is software that converts a video shot in one aspect ratio, usually widescreen, into another such as vertical or square, by tracking a subject and cropping around it automatically instead of a person repositioning every frame.
Can auto reframe handle two people in the same shot?
It can run, but it has to choose one of them to track, and that choice is a guess. Two separate accounts described the same result: the software picked a speaker and the other one's head was cropped out.
Why does auto reframe crop out someone's head?
Because it tracks a single subject, and with more than one person in frame it has no way to know which one the video is about. It keeps whoever it locked onto and crops around everyone else.
Do you still need to fix auto reframe crops by hand?
For a single centred speaker, usually not. For two or more speakers, or a speaker who moves, plan on checking the crop, since that is where the automatic guess is most likely to need correcting.