Make an On-Screen Text Conversation Work as a Commercial Scene

"Sam texts Alex about the umbrella" is not a scene. It is a note that a scene is missing.

The smallest useful correction is this: a message is four events, not one. Someone composes it. Someone sends it. A device receives it. A person reads it. A treatment can stage any of the four, hide any of the four, and cut between them wherever the pressure is highest. Messages are worth building a commercial around precisely because those events land on different people at different times. Most of the drama lives in one particular gap — between what a character typed and what the other character ever saw.

So the question that keeps a message beat honest is not "what does the phone say?" It is "who knows this, and when did they learn it?" Answer that on every beat and the interface becomes a production decision. Skip it and you have a caption with a battery indicator.

What follows is editorial judgment about staging, not a report of inspected production practice. Nothing described here has been shot, timed, or tested with an audience.

A message is four events

Composition and sending can look identical on screen — same thumb, same glow — but they do opposite work. Composition is private. Until the send control is pressed, the words exist on one device and, if the camera is pointed there, for the viewer. Sending makes them available to a second device. That is all it does. It does not make them available to a second person.

Receipt and reading are where treatments get sloppy, because both are invisible unless you build something to show them. A message arriving at a device is not a person looking at it. A notification chime tells the recipient that something arrived; it tells the sender nothing. A delivery indicator is a claim about a network, not about attention. A "seen" mark tells you a thread was opened, not what anyone understood, felt, or intends.

Which is why you check before you write. If a treatment names a platform and specifies its ticks, dots, or timestamps, that specification has to match how the software currently behaves — and these details move. Writing a state you have not verified is not the same as inventing a state on purpose. If the interface in your scene is fictional, say so in the document, and then hold every later beat to the rules you drew. A drawn interface that quietly gains a capability in scene four because the scene needs one is the same failure as a character who knows something nobody told them.

The example, and its rules

The scene below is invented. Alex lent Sam an umbrella two weeks ago and has mentioned it twice since. Sam broke it — the shaft snapped — and pushed it behind the desk in the corner, out of sight. Alex leaves the office at four for an appointment across town and wants the umbrella back before he goes. Sam knows both of those things. He would rather not be the person who has broken a third borrowed thing this quarter, and that is the whole reason the confession has to be typed instead of said.

The interface is deliberately thin, and the thinness is the point. It shows a thread of plain-text messages, a field to type in, and a send control. It does not show a typing indicator. It does not show a read receipt or a seen mark. It does not show delivery ticks. It does not show timestamps. It has no unsend, no editing, no reactions. Typed characters appear in the field; backspace removes them; pressing send commits the line and it cannot be recalled. The recipient's phone makes a notification sound on arrival. The sender's phone gives no signal at all.

Write a list like that into the treatment document. It looks fussy. It prevents the most common continuity failure in message scenes, which is a character answering something they could not have received.

The viewer's knowledge is a separate track

We watch Sam type "I broke it." We watch him erase it and send "Can we meet downstairs?" Alex's thread never contains the confession, at any point, in any state.

By the end of that beat the audience is carrying a fact available to exactly one character, and the rest of the scene runs on that debt. This is ordinary dramatic irony with an unusual delivery mechanism. The audience is not a witness to a conversation between two people; they are inside one phone.

The rule that follows is blunt: the audience's knowledge is not a channel. Whatever the viewer saw on Sam's screen did not travel to Alex's screen. Everything the second character does after that must be motivated by what the thread actually contains.

Once you accept the rule, the scene gets easier to build. Alex's reply is not a reaction to a broken umbrella. It is a request for an object he believes is intact, made by a man who is leaving the building in twenty minutes. Sam's next move is not a reaction to forgiveness or anger. It is a reaction to being asked, in plain text, for the thing he has hidden behind a desk.

Deletion and delay need a turn

A deleted draft and a delayed reply are both withheld information, and both are worth staging only if something later turns on them.

The deletion works like this: the viewer reads "I broke it," then watches it disappear, then reads what was actually sent. That is a setup, and the viewer now holds a comparison — the sentence they saw and the sentence that arrived. Sam sending a meeting request instead confirms the hiding. What complicates it is Alex's reply, because "Bring the umbrella" converts a vague meeting into a demand for the specific object Sam damaged. That is the turn. The abandoned draft pays off not in a confession but in a physical errand.

Delay works the same way and is easier to get wrong. After Sam sends, nothing comes back — no receipt, no dots, no reply. Within these drawn rules, silence carries no information about Alex's attention, and the viewer is in the same position as Sam. So the treatment cannot claim Alex is ignoring him unless the scene shows evidence: we cut to Alex on a call, or leaving the building without his phone. If you want the audience to read distraction or hostility, stage it. If you don't, let the silence stay ambiguous, which is usually more useful, because ambiguity lets the next beat land either way.

Useful ambiguity still needs a clear sequence of available information. The audience can be unsure what Alex feels. The audience should never be unsure what Alex has seen. Uncertainty about knowledge reads as a continuity error; uncertainty about feeling reads as life.

One practical consequence: if your drawn interface has no timestamps, staging has to carry elapsed time. Sam walks to the copy room. The light through the blinds has moved. A meeting lets out in the corridor. Write how much time passes, or an audience will assume the reply came instantly.

Which words have to be read

Attention is a budget, and text and performance draw on the same account. When a face is doing something and a line is appearing at the same time, the audience is choosing, and most of them will choose one.

So decide, per beat, which strings must land and which are atmosphere. In the umbrella scene, three matter: "I broke it," because it is the confession; "Can we meet downstairs?", because it is the substitution; and "Bring the umbrella," because it is the turn. The contact name at the top of the thread, the keyboard, the wallpaper, the battery icon — set dressing, and worth treating as such.

"I broke it" is the hard one, because it exists only during composition and is then erased. The audience has to read three words and register a deletion in the same short beat, which is two jobs in one moment. There are ways to buy room — hold the composer longer than feels natural, place the typed line where it sits large and high-contrast, or erase it a word at a time so the full sentence stays visible through part of the erasure. Each is a staging choice with a cost, and each is a guess until it is checked.

"Can we meet downstairs?" is easier, because it can sit on Alex's screen while something else happens. But be aware that if it arrives during a performance beat, viewers may read the notification and miss the face.

Put a target in the treatment — "the composer holds for one breath," "two seconds in the animatic" — and then test it, watching the cut with a stopwatch and someone who has not read the script. No timing test of any kind was performed for the scene above. Any number you write is a target to check, not a finding. And avoid universal reading-speed formulas: recognizing a familiar three-word phrase and parsing a new sentence are not the same task, and larger lettering does not repair a cut that is simply too short.

Finally, plan a second route for any word the scene cannot survive without — a spoken line, a visible action, an object that carries the meaning on its own. The beat-level decision about which words are load-bearing is the raw material for accessibility work, not a substitute for it.

The scene ends outside the bubble

A message scene is not over when the text is delivered. It is over when something changes between the people, and that change is almost always physical. Ask what the receiver does next that they were not going to do before.

Give the performer a job rather than an emotion. "Sam looks worried" gives the actor nothing to play. "Sam reads it, then looks at the corner behind the desk" gives them a thought and a movement, and the audience can see the thought happen.

Where does the product go? If the brand is the device or the messaging app, every control you draw is a claim about it — so either get the interface verified against the current software or keep the drawn screens generic and label them as treatment mocks. If the brand is something else, the exchange is a delivery system, and whatever you are selling should be visibly doing something in the scene rather than floating above it. Do not invent a feature because your scene wants one.

And keep approved copy where it is. If a client has signed off on a line, the treatment's job is to stage that line, not to improve it. If you genuinely need different words, flag the change in the document rather than slipping a variant in and hoping nobody compares.

The same exchange, beat by beat

The viewer's column is the table plus everything the camera saw privately, which in this scene is all of it.

Three reading moments are load-bearing: beat 1, beat 3, beat 7. Everything else can be seen without being read.

Here is the same scene with the mistake in it. Alex replies: "Don't worry about it — how bad is it?" That line is impossible. Nothing in Alex's thread contains the damage, and no other channel has delivered it. The audience knows, Alex does not, and the distance between those two facts is the entire scene.

If you want Alex to know, show how he came to know. If you give the drawn interface a typing indicator instead, Alex learns that Sam started a message and stopped — not one word of what it said. That is a different scene with a different payoff, and it is legitimate as long as you write the version you drew.

What it looks like in the treatment

Sam's desk, late afternoon. His phone. The composer shows three words: I broke it. His thumb holds the delete key and the line shortens to nothing. He types four more words and sends them. The thread now holds one outgoing message: Can we meet downstairs? He sets the phone face-down. Behind the desk, in the corner, the umbrella leans against the wall, shaft snapped.

The phone stays face-down. Sam at the printer. Sam at the window. No signal from the screen in either direction.

The phone chimes. Sam turns it over. On screen: Bring the umbrella.

He reads it. He looks at the corner behind the desk. He stands, lifts the umbrella out, and slides it into his bag, handle up.

Three readable strings, one deletion, and a last action that happens outside the bubble. Alex's reply presumes nothing he does not know. Sam's response is not a message.

The audience read the confession three beats ago, so they know exactly what is in that bag when Sam walks out — and Alex does not. That gap is the reason to put a text conversation in a commercial at all.

Frequently asked questions

What are the four events in a message, and why does that matter for a commercial scene?

A message is composition, sending, receipt, and reading. Those events land on different people at different times, so a treatment can stage any of them, hide any of them, and cut between them where the pressure is highest. The drama often lives in the gap between what a character typed and what the other character ever saw. The question to keep on every beat is who knows this, and when did they learn it.

What interface rules does the umbrella example draw, and why write them down?

It shows a plain-text thread, a field to type in, and a send control. It does not show a typing indicator, read receipt or seen mark, delivery ticks, timestamps, unsend, editing, or reactions. Typed characters appear in the field; backspace removes them; send commits the line and it cannot be recalled. The recipient's phone makes a notification sound on arrival, while the sender's phone gives no signal at all. Writing that list into the treatment prevents the common continuity failure of a character answering something they could not have received.

How does the viewer's knowledge differ from Alex's, and what rule follows?

The viewer watches Sam type 'I broke it,' erase it, and send 'Can we meet downstairs?' Alex's thread never contains the confession at any point. The audience's knowledge is not a channel: what the viewer saw on Sam's screen did not travel to Alex's screen. Everything Alex does afterward must be motivated by what his thread actually contains. If you want Alex to know, show how he came to know. Giving the interface a typing indicator would let Alex learn only that Sam started a message and stopped, not one word of what it said.

Why do deletion and delay need a turn, and what kind of ambiguity is useful?

A deleted draft and a delayed reply are both withheld information, and both are worth staging only if something later turns on them. The deletion pays off not in a confession but in Alex's 'Bring the umbrella,' which converts a vague meeting into a demand for the specific object Sam damaged, making the payoff a physical errand. After Sam sends, silence with no receipt, dots, or reply carries no information about Alex's attention. The treatment cannot claim Alex is ignoring him unless the scene shows evidence, such as Alex on a call or leaving without his phone. The audience can be unsure what Alex feels, but never unsure what Alex has seen.

Which words are load-bearing, and what timing limitation does the article state?

Three strings must land: 'I broke it' as the confession, 'Can we meet downstairs?' as the substitution, and 'Bring the umbrella' as the turn. The contact name, keyboard, wallpaper, and battery icon are set dressing. 'I broke it' is hard because it exists only during composition and is then erased, so the audience must read three words and register a deletion in the same short beat. Put a target in the treatment, such as holding the composer for one breath, then test it. No timing test of any kind was performed for the scene. Avoid universal reading-speed formulas, and plan a second route for any word the scene cannot survive without.

Further reading

• Design the Silent Version Without Turning Every Sound Into a Caption — If the words on screen must also survive without sound, this sorts what the audio was doing and rebuilds the silent version by making the phone legible.
• Show a Changing Product Interface Without Advertising a Fictional Feature — Before defending a drawn screen or a trimmed step, this separates craft compression from a simplification that quietly adds a capability the real product does not have.