What holding both people in frame actually buys
A two shot is not simply an economical way to cover a scene with fewer setups. It is a statement that these two are, for the moment, one unit of attention.
Everything spatial becomes readable. Distance between them says how comfortable they are. A height difference says something about power, especially if it was arranged rather than accidental. Whether their bodies are angled toward each other or squared away says what the scene is. Which of the two is looking says who wants something.
None of that survives a cut to singles. Coverage in singles is more flexible in the edit and it silently converts a relationship into two separate performances that the audience has to bolt back together. That is often the right choice, and it is a choice.
The useful variants
- Clean two shot. Both faces visible, roughly equal weight. Neutral, and the default for people who are aligned with each other.
- Profile framing. Both in profile facing each other. Formal, symmetrical, and it reads as either confrontation or deep intimacy depending on distance.
- Over the shoulder. One subject becomes a soft foreground mass framing the other. Preserves the connection while giving one person the frame.
- Depth two shot. One near, one far, both sharp. Old-fashioned and very strong, because it holds two scales of performance at once.
- Three shot and group shot. The same reasoning, with the added problem that someone will be the odd one out, which is usually worth designing on purpose.
Prompting a two shot
Multiple people is one of the weaker areas of image and video models, and most of the fixes are compositional rather than descriptive.
Describe the pair as clearly different people before describing anything else: an older man and a young woman. Similar ages, similar hair and similar clothing invite identity blending, where features migrate between the two faces.
State the framing for both: both framed from the chest up. Left to itself a model will often frame one subject well and clip the other awkwardly.
Give the arrangement and the eyelines explicitly: sitting side by side, neither looking at the other. Eyeline is the part models get wrong most often, and it is also the part carrying the drama.
Use one light source for the pair: cold window light from the left. Naming a single direction stops the model from lighting each face on its own terms, which is what makes generated two shots look assembled.
Finally, leave space between them. Overlapping bodies are where features merge, hands multiply and clothing swaps. A small gap costs nothing dramatically and removes most of the artefacts.