Making Inner Thoughts, Narration, and Spoken Dialogue Clear in Visual Novels
First establish whether a line is spoken within the scene and who can hear it, then assign a source label and presentation cues. An original fictional scene follows a repair technician from suspicion to quiet self-talk to a direct question, demonstrating a line-by-line convention for distinguishing how text is voiced.
Introduction
To distinguish inner thoughts, narration, and dialogue in a visual novel, first define where each line comes from, whether it is spoken within the scene, and who can hear it. Then decide how to use the name field and text styling. When a thought becomes spoken language, split the text before the first word actually spoken and signal the new type. A color change alone cannot carry that information: even readers who do not distinguish the colors should know whether a companion can respond.
The fictional teaching example below features repair technician Lin Yan inspecting a component while his companion A-Cen sits across the table. The characters, script, and conventions are original demonstrations, not a real user case, empirical results, or DramaFork product features. The editorial conventions proposed here follow from whether words are voiced and who can hear them; they are not presented as a universal industry standard.
Establish whose words they are, then who can respond
For each line, the lead editor asks three questions: Whose thought or narration is this? Does it become a sound within the scene? If it does, who hears it, and which parts do they hear? Record the three answers separately. The name field cannot stand in for all of them.
“Lin Yan” only identifies whom the text relates to; it does not prove that he has opened his mouth. “Quietly” describes how he speaks, but it does not automatically mean A-Cen cannot hear him. How the distance across the table, ambient noise, and character actions affect what is heard must be established explicitly in the current scene. Readers should not have to supply conditions the author has left unstated.
Narration also requires a distinction between the narrative source and a source of sound within the scene. “There is a scratch along the edge of the component” can come from a narrator outside the scene. “That scratch reminds me of the old warehouse” contains a character’s personal experience. This example uses the former approach and limits narration to observable facts. To enter Lin Yan’s subjective judgment, switch to his inner thoughts so that the perspective does not change within a passage without warning.
Set six fields for a voicing convention
The following conventions can be used directly for this example. Text in brackets represents cues intended for readers. The scope of what can be heard is an editorial record; the whole field does not need to be shown to readers.
| Type | Source | Who can hear what in the scene | Presentation cue | How to switch | Example of misuse |
|---|---|---|---|---|---|
| Inner thought | Lin Yan’s current thoughts | No one hears it | [Lin Yan · Thought] | Start a new text unit before speech begins | A-Cen answers an unspoken thought directly |
| Narration from outside the scene | A narrator who is not a character in the scene | No one hears it | [Narration] | Change the label when moving into a character’s judgment | The narration suddenly recalls something in the first person |
| Quiet self-talk | Lin Yan actually speaking | Specify the listener and the words clearly heard for each line | [Lin Yan · Quiet self-talk], with an explicit account of what is heard when needed | Start a new unit when addressing someone | Assuming self-talk can never be overheard |
| Speech addressed to someone | The character currently speaking | Both people at the table in this example | [Lin Yan] or [A-Cen]; establish the addressee in the text | Start a new unit when switching from thought or self-talk | Using one label for both a thought and a spoken question |
To reuse this approach, copy all six fields first, then specify the sources and who can hear what for your work. Do not keep only the first three columns: the categories may remain clear, but the people producing the scene will still lack instructions for handling transitions.
Color, borders, and fonts can reinforce these cues, but this example keeps text labels as the foundation. Persistent labels add visual clutter, yet they help readers pick up the story midway through. If a work uses fewer labels, it needs another stable set of cues that can be recognized on their own. Check whether those distinctions still hold when reviewing the text in plain text.
Put three changes into practice in a complete short scene
First, fix the scene conditions: the two characters sit across a table, and A-Cen is leafing through a repair log. Lin Yan has not yet shared his suspicions with him. When Lin Yan mutters to himself, A-Cen clearly hears only “old warehouse.” This limit on what is heard is an authored condition of the example, not a conclusion inferred from “quietly.”
[Narration] Lin Yan turns the component over, resting the pad of his finger on a half-worn marking.
[Lin Yan · Thought] This marking looks like the ones on goods from the old warehouse. A-Cen said he just bought it. I should probably ask more about where it came from.
[Lin Yan · Quiet self-talk] Old warehouse… What’s it doing here?
[Narration] A-Cen looks up. The only words he caught clearly were “old warehouse.”
[A-Cen] Did you just mention the old warehouse?
[Narration] Lin Yan pushes the component to the center of the table and turns to A-Cen.
[Lin Yan] Who did you buy this component from?
[A-Cen] The vendor by the bridge. Did you find something?
Here, A-Cen first checks the words he heard, then answers the direct question. He does not respond to “I should probably ask more about where it came from,” because that thought is available only to the reader. Lin Yan’s actions—pushing the component forward and turning toward his companion—help establish that he is about to address him.
If the story requires A-Cen to miss the self-talk entirely, change the account of what he hears to “A-Cen is still leafing through the log and does not hear the muttered words,” and remove his question confirming what he heard. This places more emphasis on Lin Yan’s private suspicion, but loses the turn in which his companion notices something is amiss. Whenever you change who hears what, revise the subsequent responses as well.
Split a mid-sentence transition before the first spoken word
An easily misread draft is: “[Lin Yan · Thought] Could this marking be from the old warehouse… where did you buy it?” The author may intend the second part to slip out aloud, but the label still defines the entire line as thought. An ellipsis indicates a pause or interruption; it cannot establish a change from thought to speech on its own.
If speech is meant to begin with “old warehouse,” make two passages: “[Lin Yan · Thought] Could this marking be from the—” followed by “[Lin Yan · Quiet self-talk] Old warehouse?” If only the final question is spoken, keep everything before it as thought and begin a new unit of speech addressed to someone before “where.” The split depends on the first word actually voiced, not on where a full stop falls.
There is no need to squeeze two types back into one text unit just to preserve complete grammar. An unfinished sentence can retain the feeling of an interrupted train of thought; a complete sentence gives a steadier rhythm. Whichever rhythm you choose, the new label should appear before the text of the new type, so readers do not discover only afterward that they interpreted it incorrectly.
Check the limits of character knowledge through misplaced responses
Before handing over the script, add an internal note for every response: “Which piece of information is this responding to, and how was it obtained?” In this example, A-Cen asks about the old warehouse because he clearly heard those words. He explains where he bought the component because he heard Lin Yan’s direct question. Both have a basis in the text.
A counterexample would have A-Cen immediately follow the inner thought with “I didn’t lie to you. I really did just buy it.” This directly disputes a suspicion that has not been shared. If the author wants him to infer doubt from Lin Yan’s behavior, provide visible evidence, such as Lin Yan repeatedly examining the marking, and make the response tentative: “Do you think there’s something wrong with it?” A guess should not be written as if the character accurately heard someone else’s thoughts.
During the check, temporarily remove colors and staging effects, and read only the labels, text, and responses. If it is still unclear whether a line is spoken, add a textual cue. If a character’s response exceeds what they know, revise the response or add an actual transfer of information within the scene. Do not cover the gap with an explanation after the fact.
These conventions suit passages that need to make communication within a scene clear. Dreams, telepathy, or stories that deliberately obscure the source of a voice can use different rules, but record where each exception begins and who receives what information. Narration whose source has not yet been revealed can temporarily use [Narration · Source unrevealed], with its narrative boundaries fixed in the editorial script. Use [Unknown voice] only for a voice that actually sounds within the scene. Concealing an identity still requires preserving the distinction between voiced and unvoiced text.
