The third person is the scene
A confession between two people is a romance. A confession that a third person watches from the doorway is a triangle, and the whole charge of the scene sits in that third face. The studio allows three cast members per take, each one passed to the model as a face reference with its own costume and prop, so you can put the watcher in the frame rather than describing them in a sentence and hoping.
The school romance tone helps here. Its expression vocabulary includes eyes that widen then drop, a hand tightening on a bag strap and a smile that arrives late, which is the vocabulary of someone pretending not to mind. Write what the third person does with their hands and the drawing will do the rest.
Tropes that hold a triangle
One umbrella in the rain is the obvious one: a downpour at the campus gate, one clear umbrella, two people under it and a third checking the sky alone. The three beats take the pair to the crossing, where one of them says the thing the rain gave them cover to say; where you place the third character is up to you. The rooftop lunch works as the aftermath, three lunchboxes instead of two and a pause too long to be nothing.
For the sharper edge of a triangle, switch genre to rivalry and use the handshake refused. Two people who want the same thing, a hand offered, a choice shown in whether it is taken, with the person they both want standing just out of focus.
- Shared umbrella: inside and outside the umbrella are two different scenes. Render both.
- Rooftop lunch: three lunchboxes; who brought the extra one is the story.
- Handshake refused (rivalry): the triangle as a contest, with the camera on whoever flinches.
Casting three people
The preset library has sixteen adults with distinct silhouettes so three of them read clearly in a vertical frame. Kaede Arisugawa with her chestnut bob, Shunpei Hoshizaki with spiky black hair and a scarred eyebrow, and Nanami Tsukishiro with silver-lilac twin tails are hard to confuse even at phone size. Give each a different outfit, keep props to one or two so the reference count stays under nine, and pick one set for all of them.
Voice styles pull the triangle apart nicely: Sincere for the one who means it, Teasing for the one who deflects, and Dry wit for the one left holding the umbrella. These are written delivery directions, not recordings, and they are never anybody’s actual voice.
Writing the scene so three people are legible
Name all three using their exact library names, and give each one a single action. One holds the umbrella. One steps under it. One stops at the gate and says nothing. Then one spoken line, in quotation marks, from whoever the scene belongs to. Avoid overlapping dialogue; a take performs lines aloud and three voices in eight seconds turns into noise. If two people need to speak, give the third a look instead.
Keep the take between six and ten seconds. Shorter and the third person has no time to register; longer and the model starts inventing a fourth.
What a triangle costs
Three characters cost the same as one: takes are priced by length, quality and resolution, not by cast size. An eight-second standard take at 720p is forty-eight credits whether one person or three is in it. The price appears in the studio before you render, and credits are returned only if the take fails before rendering starts.