How to Generate Phone Screens and Text Messages Inside AI Vertical Drama Shots So They Stay Readable

The message arrives. The lead looks down at her phone. The camera pushes in on the screen, and the screen shows a smear of letter shaped marks that change every frame. The line that was supposed to land the cliffhanger, the one that reveals who sent it, cannot be read. The episode still uploads, but the hook that was meant to drive the unlock does not work, because the audience never saw it.

Phone screens are everywhere in vertical drama. Texts from a secret lover. A bank alert. A photo that changes everything. A missed call from a name that should not be calling. The genre relies on them because they deliver plot information quickly and intimately, and because viewers are themselves watching on a phone. The format practically invites the device into the story.

Generators, however, are poor at text. They produce shapes that look like writing at a glance and fall apart under inspection. Letters warp, words repeat, interfaces flicker between frames. This guide sets out a workflow that treats on screen text as a separate layer rather than something the generator is asked to produce, and covers the design, compositing, continuity and review steps that keep phone screens readable from the first episode to the last.

1. Why Generators Struggle With Readable Text

The underlying reason is that video generators model appearance, not meaning. They have learned what text tends to look like in an image, so they can produce convincing texture, but they do not hold a stable representation of individual characters across frames. Each frame is an approximation, and approximations of letters are not letters.

The problem worsens in motion. A static image generator can sometimes render a short word correctly. A video generator has to keep that word identical across every frame while the phone moves, the hand shifts and the light changes. Small variations between frames are invisible on a face but immediately visible on text, because the eye reads text as a fixed code. Any change reads as flicker.

Interfaces add another layer of difficulty. A messaging app screen contains many small elements: timestamps, contact names, bubbles, status icons, a keyboard. Generators tend to produce plausible clutter rather than a coherent layout, and the clutter shifts frame to frame. Even when the message text is close to correct, the surrounding interface betrays the shot as generated.

The practical conclusion is simple. Do not ask the generator to write the message. Ask it to produce the hand, the phone, the light and the performance, and add the screen content as a designed layer afterwards. Every other step in this guide follows from that decision.

2. Decide How Each Phone Moment Will Be Shown

Not every phone moment needs the same treatment. Before generation, each one should be assigned a display method based on how much the audience needs to read and how long the moment lasts. Making this decision at the shot list stage saves regeneration later.

The first method is the full screen insert. The frame is filled entirely by the phone display, rendered as a designed graphic with no generated imagery at all. This is the most readable option and works well for long messages, multiple message threads or anything the audience must read carefully. It costs nothing in generation and gives complete control over typography.

The second method is the in hand screen with composite. The generator produces the character holding the phone, with the screen deliberately left as a plain, softly lit surface. The designed screen is tracked onto that surface in post. This keeps the performance and the device in the same shot and is the right choice when the audience needs to see the character react while reading.

The third method is the overlay. The character is shown in a normal shot and the message appears as a floating graphic beside or above them, styled as a message bubble but not attached to a device. Many vertical drama platforms use this convention because it keeps the face in frame while delivering the text. It is the fastest method and reads well at small sizes, though it is less naturalistic.

The fourth method is the unreadable glance. Sometimes the story only needs the audience to know a message arrived, not what it said. A quick shot of a lit screen, angled away or out of focus, is enough. In this case the generator can produce the screen glow directly, because nobody is meant to read it.

3. Design a Locked Screen System for the Series

A series with dozens of phone moments needs a consistent screen system, designed once and reused. Without one, each insert is built separately and the messaging app subtly changes across episodes, which audiences notice in the same way they notice wardrobe drift.

The screen system starts with a fictional interface. Avoid recreating any real messaging app, banking app or social platform, both to avoid brand issues and because platform content policies often discourage it. Design a clean, generic messaging layout with its own bubble shape, colour palette, header and status bar. It should feel familiar without copying anything specific.

Each principal character who sends or receives messages should have a defined device. Phone model silhouette, case colour, lock screen wallpaper and contact photo as seen by other characters. These are continuity assets in exactly the same way as wardrobe, and they should be logged in the same register so the device in episode forty matches the device in episode three.

Typography for a Phone Inside a Phone

The audience is reading a phone screen that appears inside a frame on their own phone screen. That doubling shrinks everything. Text that would be comfortable on a real device becomes tiny once it occupies a fraction of a 9:16 frame on a handheld display. The principles of legibility apply directly: clear letterforms, generous size, strong contrast and adequate spacing matter more than stylistic fidelity to a real interface.

In practice, message text in an in hand composite should be scaled well beyond what a real phone would show, so that the words remain readable on the viewer device. Use a sans serif typeface with open counters, keep line lengths short, and avoid thin weights. Dark text on light bubbles generally outperforms light text on dark bubbles at small sizes, though either works if contrast is high enough.

Write Messages for Reading Speed

Messages should be short. A viewer needs time to find the text, read it, and register its meaning before the cut. A useful rule is to hold any readable message on screen long enough to read it twice at a relaxed pace. If the message is long, split it into several bubbles that appear in sequence, each held long enough to read. The scriptwriter and the editor should agree message wording and hold durations before the shot is designed.

4. Generate the Shot With a Clean Screen Surface

For the in hand composite method, the generation prompt needs to produce a phone with a screen that is easy to replace. That means a uniform, softly lit surface without detail, at an angle that allows readable placement, and with the hand positioned so fingers do not cross the screen area more than necessary.

Prompt the screen as a plain, evenly illuminated display, describing the light it casts on the face rather than the content it shows. Many generators will still try to add interface clutter, so a negative instruction against text, icons and app interfaces helps where the tool supports it. Even partial clutter is acceptable if it is low contrast, because the composite will cover it.

Angle matters more than teams expect. A phone held flat to camera is easiest to composite but can look staged. A phone tilted slightly away is more natural but compresses the text. A good compromise is a gentle tilt of the device toward camera, enough to feel natural while keeping the screen surface broad enough for readable text. Specify the angle in the prompt and in the shot plan so the editor knows what to expect.

Keep the Screen Light Consistent With the Face

The screen should light the face. A character reading a message in a dark room needs a cool glow on the cheekbones and under the eyes, and that glow should be consistent with the screen brightness that will be composited later. If the generated face lighting suggests a bright white screen but the composite shows a dark mode interface, the shot will look wrong even if the viewer cannot say why. Decide the screen brightness and colour temperature in pre production and prompt the face light to match.

5. Composite the Screen So It Sits in the Shot

Once the shot is generated, the designed screen is placed onto the device surface in post. This is standard screen replacement work, and the low cost compositing stack that most vertical drama teams already use is sufficient. The aim is not technical perfection. The aim is a screen that moves with the phone, sits inside the frame of the device, and carries the same light and grain as the rest of the image.

Track the four corners of the screen across the shot. Generated footage sometimes has subtle warping in the device itself, so check that the tracked screen does not slide against the bezel. Where the phone geometry wobbles, a stabilised composite with slight manual correction is usually enough to hold the illusion.

Match the image characteristics. Add the same grain or noise as the generated plate. Apply a slight softness so the screen does not look sharper than the hand holding it. Add a faint reflection or glare that moves with the device angle. Where fingers cross the screen, rotoscope them in front of the composite. These details are small individually and decisive together.

Animate Messages, Do Not Just Place Them

A message arriving should behave like a message arriving. A bubble appearing with a short animation, a typing indicator before a reply, a notification banner sliding down. These behaviours are familiar to every viewer and make the screen feel alive. They also give the editor control over timing, because the moment the text appears can be placed precisely on the beat of the performance.

6. Protect Readability Against Platform Interface Overlays

Vertical drama platforms place their own interface over the video: episode titles, progress bars, unlock buttons, like and share icons, subtitle tracks. A message composited in the wrong part of the frame can be hidden by any of these. Before designing screen positions, the team should know the safe zones for every platform the series will appear on.

As a general principle, keep readable phone content in the central band of the frame, away from the top edge where titles sit and the bottom third where controls and subtitles cluster. Avoid the right edge, where many platforms stack engagement icons. If the platform list is not known at production time, design to the most restrictive common layout.

Subtitles deserve special attention. If dialogue subtitles run in the lower third and a message is also on screen, the two compete for attention. Either pause dialogue while a message is read, or position the message so it is clearly separate from the subtitle line. Where the message itself is the line, consider whether it should also appear as a subtitle for accessibility, styled differently from spoken dialogue.

7. Build Phone Screens for Localisation From the Start

Phone screens contain written language. If the series will be delivered in more than one language market, every message needs to be translatable without regenerating the shot. The layered workflow makes this possible, but only if the layers are kept separate and documented.

Keep each screen design as an editable project with live text, not a flattened image. Store the message copy in a single document indexed by episode, shot and bubble, so translators can work from one source. Allow space in the bubble design for longer translations, because many languages run significantly longer than English. A bubble that fits a short English word may overflow in another language.

Character names, contact labels and timestamps also need localising. A contact saved as a nickname in English may need a different nickname in another market. Timestamps should follow local conventions. These details are minor, but a localised series that keeps English interface elements on its phone screens looks unfinished to the audience it is trying to reach.

Review Every Screen at Viewing Size

The final check for any phone moment is to watch it on a phone, at normal viewing distance, at the speed it will play. Reviewing on a large monitor hides problems, because text that is easily readable at desktop scale may be unreadable on a handheld screen. Every phone insert should be signed off by a reviewer who has watched it on a device, read the message comfortably, and confirmed it sits clear of platform overlays.

Axis AI Studios Perspective

Axis AI Studios produces AI native vertical drama, and phone screens are planned as designed layers from the shot list onward. Each phone moment is assigned a display method before generation. Each series gets a locked fictional interface and a device register for its principal characters. Shots are generated with clean screen surfaces and matching face light, screens are composited and animated in post, and every insert is reviewed on a handheld device against platform safe zones.

The result is the reveal landing where it should. The audience reads the message, understands who sent it, and taps through to the next episode. That is the whole job of a phone screen in this format, and it is worth building a workflow around.

Platforms, brands and media companies commissioning AI vertical drama who want to talk through how on screen text and phone moments would be handled on their series can reach Axis AI Studios at business@axisaistudios.com.

FAQ

Can a video generator produce readable text messages directly?

Occasionally for a single short word in a static shot, but not reliably across frames in motion. The dependable approach is to generate the phone with a plain screen surface and composite the designed message in post, or to use a full screen graphic insert.

Should phone interfaces copy a real messaging app so they feel familiar?

No. A generic fictional interface feels familiar enough and avoids brand and platform content policy issues. Design it once for the series and reuse it so it stays consistent across every episode.

How long should a text message stay on screen?

Long enough for a viewer to read it twice at a relaxed pace. Longer messages should be split into several bubbles that appear in sequence, each held for its own reading time, with hold durations agreed between writer and editor before the shot is designed.

Further Reading

For designers working out how on screen text should be sized and positioned for handheld viewing, the guide to lower third design for vertical drama mobile legibility covers typography, contrast and placement principles that carry straight over to phone screens.

For writers and editors relying on messages to carry a hook when sound is off, the piece on writing a hook that works without sound covers how visual information carries an opening under muted autoplay.

For post teams setting up the compositing work that screen replacement depends on, the overview of the low cost vertical drama VFX stack covers the tools and techniques that make tracked inserts practical at series volume.