Dankoma

GitHub Repository · Documentation

High-performance Bilibili-style danmaku renderer with Mode 7 support.

Dankoma is a JavaScript danmaku renderer built on HTML Canvas. It supports scrolling, fixed-position, reverse-scrolling, and Mode 7 danmaku, with timeline-based playback, JSONL loading, collision management, and rendering caches.

Features

• Bilibili-style danmaku rendering

• Scrolling, top-fixed, bottom-fixed, and reverse-scrolling modes

• Mode 7 danmaku with:

  • Positioning
  • Movement
  • Easing
  • Rotation
  • Perspective transformation
  • Opacity animation
  • Font and outline configuration

• Video timeline synchronization

• JSONL loading and parsing

• Blob support for JSONL loading

• Cached text metrics and rendered sprites

• Cached Mode 7 transformations

• Lane-based collision management

• Device-pixel-ratio configuration

• OffscreenCanvas support when available

Installation

Dankoma can be used directly in a browser environment.

<canvas id="danmaku"></canvas>
<video></video>

<script src="https://cdn.jsdelivr.net/npm/dankoma.js@0.1.3"></script>

Basic Usage

const canvas = document.getElementById("danmaku");
const video = document.querySelector("video");

const dankoma = new Dankoma(canvas, {
    style: {
        opacity: 0.8,
    },

    scroll: {
        duration: 6.5,
        lookahead: 8,
        gap: 2,
    },

    fixed: {
        lifetime: 5000,
    },
});

dankoma.trackVideo(video);

await dankoma.loadDanmaJSONL("/danmaku.jsonl");

Once the video is playing, Dankoma follows its playback position and schedules the corresponding danmaku.

JSONL Format

Dankoma accepts one danmaku record per line:

["Hello world",1.0,1,0,16777215,500]
["Nice song",2.5,1,0,16776960,500]
["Top comment",4.0,5,0,16777215,500]
["Bottom comment",6.0,4,0,16777215,500]

The record format is:

[text, time, mode, timestamp, color, weight]

Where mode determines the rendering behavior:

Mode	Description
1	Scrolling
2	Scrolling
3	Scrolling
4	Bottom fixed
5	Top fixed
6	Reverse scrolling
7	Mode 7

Mode 7

Mode 7 records use an extended JSON array payload.

The payload format is:

[x1, y1, opacity, lifetime, text, zRotation, yRotation,
 x2, y2, moveDuration, delay, outline, fontFamily, linear]

For example:

[
    0.5,
    0.5,
    "1-0",
    4,
    "Hello Mode 7",
    0,
    20,
    0.8,
    0.5,
    2000,
    0,
    1,
    "Noto Sans",
    0
]

As a complete danmaku record:

[
    "[0.5,0.5,\"1-0\",4,\"Hello Mode 7\",0,20,0.8,0.5,2000,0,1,\"Noto Sans\",0]",
    8.0,
    7,
    0,
    16777215,
    32
]

Mode 7 supports positioning, movement, easing, rotation, perspective effects, opacity animation, fonts, and outlines.

See the Mode 7 specification for the complete format and supported fields.

Loading Danmaku

JSONL data can be loaded from a URL or Blob:

await dankoma.loadDanmaJSONL("/danmaku.jsonl");

loadDanmaJSONL() appends records to the existing danmaku dataset rather than replacing it.

The source may be either a URL accepted by fetch() or a Blob:

const blob = new Blob([
    '["Hello world",1.0,1,0,16777215,500]\n'
], {
    type: "application/jsonl",
});

await dankoma.loadDanmaJSONL(blob);

When a Blob is supplied, Dankoma creates a temporary object URL internally and loads the JSONL from it.

For manually adding individual records, use:

dankoma.appendDanmaku([
    "Hello world",
    10.0,
    1,
    0,
    16777215,
    500,
]);

For lower-level stream handling, readTextStream() and parseJSONL() are also available:

const text = await dankoma.readTextStream(stream);

dankoma.parseJSONL(text, comment => {
    dankoma.appendDanmaku(comment);
});

The stream is decoded first, then the resulting text is parsed as JSONL.

Runtime API

Video

dankoma.trackVideo(video);
dankoma.untrackVideo();

Associates or removes the video used for timeline synchronization.

Visibility

dankoma.hide();
dankoma.unhide();

Temporarily disables or re-enables rendering.

Danmaku Data

dankoma.appendDanmaku(danmaku);

dankoma.clearDanmakus();
dankoma.resetDanmakuData();

appendDanmaku() adds a single record to the loaded dataset.

clearDanmakus() removes currently active danmaku while retaining the loaded dataset.

resetDanmakuData() removes the loaded danmaku data and resets the playback timeline.

Configuration

dankoma.updateConfig({
    style: {
        opacity: 0.9,
    },
});

Updates the renderer configuration.

Cleanup

dankoma.destroy();

Stops rendering and releases the renderer's resources.

Rendering

Dankoma separates danmaku scheduling, state updates, and rendering.

Rendered text and Mode 7 transformations are cached where possible, while active danmaku are updated according to the current playback position.

The renderer also maintains lane state for scrolling and fixed-position danmaku to manage overlapping comments.

Documentation

For the complete API reference, configuration details, JSONL format, Mode 7 specification, internal functions, and implementation notes:

Documentation

License

See the repository's license file for licensing information.