Where the frame line sits
The waist is the conventional cut, but the useful definition is behavioural rather than anatomical. This is the size at which you can see what someone's face is doing and what their hands are doing at the same time. Cut higher and the hands leave. Cut lower and the face stops carrying the scene. Everything else about the size follows from that one property.
In practice the line drifts with the number of people in frame. A two-person medium is wider than a single medium, because two bodies need more horizontal room, and a three-person version starts overlapping with a medium wide. Nobody renames it, and the shot list still says "MS".
Why dialogue coverage lives here
Most scenes are two people talking, and most of what a performer does while talking happens between the chin and the belt: a shrug, a hand on a counter, a shift of weight. A close-up throws all of that away in exchange for a bigger face, which is worth it exactly once per scene. A wide keeps the body but makes the face too small to read.
There is also an editing reason. A two-person medium can hold a full exchange without a cut, so an editor gets a choice: stay in the wider frame for rhythm, or cut to singles for pressure. Coverage that only exists in singles removes that option.
The variants worth knowing
- Medium wide shot. Cuts near the knee. Keeps a sense of stance and full-body movement, sometimes called a cowboy shot.
- Medium shot. Waist up. The neutral version.
- Medium close-up. Chest up. The bridge into a close-up, and the most common size for interviews and news.
- Medium two shot. Two people, waist up, sharing the frame. Reads as connection when they are close and as distance when the composition puts air between them.
- Dirty medium. A single medium with a piece of the other character breaking the foreground edge. Keeps the second person present without giving them the frame.
When it is the wrong choice
This size has one real weakness: it is neutral, and neutral is not always what a scene needs. Holding a medium through a moment that turns on a flicker of doubt wastes the moment, because the face is competing with a jacket, a shoulder and half a room.
It is also a poor opening size. An audience that does not yet know where they are will spend the first seconds of a medium trying to read the background instead of listening. Give them a wide shot or an establishing shot first, then the medium works immediately.
Prompting one
The label itself is well understood by image and video models, which makes this one of the more reliable sizes to request. Two things still go wrong: the frame drifts wider than asked, and the subject ends up centred and static.
- Reliable:
medium shot, waist up framing - Add a redundant cue when it drifts:
medium shot, subject framed from the waist up, hands visible at the bottom of frame - Weak on its own:
mid shot, which models often read as a generic portrait
Give the size a lens and a subject distance, since medium shot, 50mm, camera about two metres from the subject pins perspective as well as crop. If a person keeps coming back full body, name what should be outside the frame, for example legs not visible. And if you want the frame to feel like coverage rather than a portrait, ask for the subject to be turned slightly off axis and looking out of frame, because a straight-to-lens medium reads as an interview no matter what else the prompt says.