Danmu (弹幕) are comments attached to a particular moment in a video and displayed over the image during playback. The writer may have watched yesterday and the next viewer may arrive next month, but both comments appear at the relevant second. The interface turns asynchronous posting into the feeling of a shared reaction.

That time attachment—not the fact that text moves—is the core mechanism. It changes what a comment is for, how it is read and what can go wrong when too many people write at once.

A comment belongs to a moment

In a conventional comment section, messages sit below the video and are usually ordered by posting time, popularity or replies. To understand one, a reader may need to remember the whole video or follow a thread.

Danmu are indexed to playback time. A warning can appear just before an event. A joke can land on the frame that makes it funny. A repeated phrase can build as the same character enters or the same musical cue returns.

Bilibili’s annual filing defines its “bullet chat” as a function in which comments fly across the screen and are contextual to the content being watched. An independent open-access study describes the same structure: comments are overlaid and synchronised to the point at which users sent them.

The audience is together—and not together

The interface produces what researchers call a pseudo-synchronous effect. Viewers experience reactions at the same narrative moment, even though the people who wrote them were not necessarily online together.

This creates a distinctive form of address. A person can write “look at the left” without naming another user because the video frame supplies the context. Others can repeat a phrase without using a formal reply button. The screen itself becomes the shared thread.

The illusion has limits. A danmu does not prove live consensus, and a dense wave of identical text may come from imitation rather than independent agreement. The interface aligns timing, not identities or motives.

Why “bullet curtain” is a useful translation

The Chinese characters 弹幕 literally refer to a barrage or curtain of fire. On video, many comments moving together can form a visual curtain that partially covers the image. The name describes the appearance at high density.

“Bullet comments” and “bullet chat” are common English renderings. Neither means the messages are necessarily aggressive. They may be jokes, explanations, lyrics, warnings, emotional reactions or repeated community phrases.

The metaphor also points to the main usability cost: the social layer can overwhelm the work it is discussing.

Density is both the feature and the failure mode

One comment placed at the right second can add context. Hundreds can obscure faces, subtitles or interface elements. The same accumulation that signals a shared moment can make the underlying video hard to watch.

Platforms therefore provide controls such as switching danmu off, filtering content or changing display density. Those controls matter for accessibility as well as preference. Moving text can compete with subtitles, increase visual load and create problems for viewers who need a stable image.

Turning danmu off does not mean a viewer has misunderstood the culture. It is an interface choice, like muting a group chat while continuing to read selected messages later.

Danmu are not subtitles

Both appear over video, but they have different authority. Subtitles normally represent dialogue, translation or editorial description. Danmu are user messages. They can be wrong, ironic, late, deliberately misleading or dependent on an in-group reference.

That distinction is especially important for an overseas viewer using machine translation. A sentence over a person’s face may look like a caption while actually being a viewer’s joke. Position and timing do not make it part of the original work.

The interface shapes community without guaranteeing one

The academic study in Media and Communication argues that users build interaction despite weak address information and limited thread structure. Repetition, timing and reference to the video help produce order from a visually loose stream.

That is a mechanism, not a claim that every danmu space is welcoming. Shared codes can create belonging for insiders and confusion for newcomers. Anonymity can lower the cost of participation and the cost of harassment at the same time. Moderation rules and platform governance remain part of the experience.

How to read danmu as cultural evidence

Treat the stream as a situated audience layer, not a survey of public opinion.

  • Check the timestamp. What exact frame or sound is the comment responding to?
  • Look for repetition. A phrase may be a platform convention or reference, not a literal statement.
  • Separate overlay from source. The video creator and the commenter are different speakers.
  • Keep scale honest. Comments on one video describe one participating audience, not “Chinese people.”
  • Use the controls. Pause, reduce density or turn the layer off when it prevents comprehension.

Danmu are valuable because they preserve reactions at the moment they make sense. They are limited because that preservation removes much of the normal context about who spoke and why.

Sources and scope

The article explains an interface and its documented social affordances. It does not claim that danmu originated in China, that every Chinese viewer uses it or that visible comment density measures national opinion.