### Install and Run Development Server

Source: https://github.com/doedja/scenecut/blob/main/README.md

Install dependencies and start the development server for the browser application. The `build:core:wasm` command requires the emscripten SDK.

```bash
bun install
bun run build:core:wasm        # needs emscripten SDK
bun run dev:app
```

--------------------------------

### Install Dependencies and Build Project

Source: https://github.com/doedja/scenecut/blob/main/README.md

Installs project dependencies using Bun and then builds the core components (WASM, TS), Node.js CLI, web client, and application.

```bash
bun install

# one shot: builds core (WASM + TS), node, web, app
bun run build
```

--------------------------------

### Install @doedja/scenecut-web

Source: https://github.com/doedja/scenecut/blob/main/packages/web/README.md

Install the package using npm or yarn.

```bash
npm install @doedja/scenecut-web
```

--------------------------------

### Install @doedja/scenecut CLI

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Install the scenecut command-line tool globally using npm.

```bash
npm install -g @doedja/scenecut
```

--------------------------------

### Quick Start with Vite

Source: https://github.com/doedja/scenecut/blob/main/packages/web/README.md

Initialize WASM factory and motion pool, then use detectSceneChanges for video analysis. Ensure WASM files are served as static assets.

```typescript
import {
  detectSceneChanges,
  createWebWasmFactory,
  createWebMotionPool
} from '@doedja/scenecut-web';

// Serve these from your public folder (copy from node_modules/@doedja/scenecut-core/dist/)
const wasmBase = new URL('/wasm/', location.href).href;
const glueUrl  = `${wasmBase}detection.wasm.js`;
const wasmUrl  = `${wasmBase}detection.wasm`;

const wasmFactory = createWebWasmFactory({ glueUrl, wasmUrl });

const pool = createWebMotionPool({
  glueUrl,
  wasmUrl,
  createWorker: () => new Worker(
    new URL('./my-motion-worker.ts', import.meta.url),
    { type: 'module' }
  )
});

async function detect(file: File) {
  const result = await detectSceneChanges(file, {
    wasmFactory,
    pool,
    sensitivity: 'low',
    onProgress: (p) => console.log(`${p.percent}%`),
    onScene: (s) => console.log(`cut @ ${s.timecode}`)
  });
  return result.scenes;
}
```

--------------------------------

### scenecut CLI Usage Example

Source: https://github.com/doedja/scenecut/blob/main/apps/web/index.html

Demonstrates how to use the scenecut command-line interface to detect scene changes in a video file and specify the output format.

```shell
scenecut video.mkv --format edl
```

--------------------------------

### Cancellation Example

Source: https://github.com/doedja/scenecut/blob/main/packages/web/README.md

Demonstrates how to abort a scene detection process using AbortController and passing its signal to the detectSceneChanges function.

```typescript
const ctrl = new AbortController();
const result = await detectSceneChanges(file, { wasmFactory, pool, signal: ctrl.signal });
// later:
ctrl.abort();
```

--------------------------------

### Build Core WASM Module

Source: https://github.com/doedja/scenecut/blob/main/README.md

Builds the core WebAssembly module. Requires the Emscripten SDK to be installed.

```bash
bun run build:core:wasm
```

--------------------------------

### Motion Worker Setup

Source: https://github.com/doedja/scenecut/blob/main/packages/web/README.md

Configure the motion worker script to handle WASM initialization and motion detection tasks. This script is instantiated by the bundler for each worker.

```typescript
import { installMotionHandler } from '@doedja/scenecut-core';
import type { WorkerMessagePort } from '@doedja/scenecut-core';
import { createWebWasmFactory } from '@doedja/scenecut-web/wasm-factory';

installMotionHandler(self as unknown as WorkerMessagePort, (glueUrl, wasmUrl) =>
  createWebWasmFactory({ glueUrl, wasmUrl })
);
```

--------------------------------

### Browser Library Usage for Scene Detection

Source: https://github.com/doedja/scenecut/blob/main/README.md

Integrate the SceneCut browser library to detect scene changes in videos. This example shows how to set up the WebAssembly factory and motion pool, and then call the `detectSceneChanges` function.

```javascript
import {
  detectSceneChanges,
  createWebWasmFactory,
  createWebMotionPool
} from '@doedja/scenecut-web';

const wasmFactory = createWebWasmFactory({
  glueUrl: new URL('./wasm/detection.wasm.js', location.href).href,
  wasmUrl: new URL('./wasm/detection.wasm', location.href).href
});

const pool = createWebMotionPool({
  glueUrl: '...', wasmUrl: '...',
  createWorker: () => new Worker(
    new URL('./motion-worker.ts', import.meta.url),
    { type: 'module' }
  )
});

const result = await detectSceneChanges(file, {
  wasmFactory,
  pool,
  sensitivity: 'low',
  onScene: (s) => console.log(s)
});
```

--------------------------------

### Build Application

Source: https://github.com/doedja/scenecut/blob/main/README.md

Builds the main application, likely encompassing all necessary components.

```bash
bun run build:app
```

--------------------------------

### Build Web Client

Source: https://github.com/doedja/scenecut/blob/main/README.md

Builds the web client components for browser-based usage.

```bash
bun run build:web
```

--------------------------------

### Build Node.js CLI

Source: https://github.com/doedja/scenecut/blob/main/README.md

Builds the Node.js command-line interface for the project.

```bash
bun run build:node
```

--------------------------------

### Generate scene thumbnails

Source: https://github.com/doedja/scenecut/blob/main/README.md

Create thumbnail images for detected scenes and save them into a specified directory.

```bash
scenecut video.mp4 --thumbnails ./thumbs
```

--------------------------------

### Adjust sensitivity and enable verbose output

Source: https://github.com/doedja/scenecut/blob/main/README.md

Run scenecut with higher sensitivity to detect more subtle transitions and enable verbose mode to see per-scene details.

```bash
scenecut anime.mkv -s high -v
```

--------------------------------

### Import WASM Binary (Vite)

Source: https://github.com/doedja/scenecut/blob/main/packages/core/README.md

Import the raw WASM binary file using Vite's URL import syntax. This is a shortcut for importing the compiled WASM module.

```typescript
import wasmUrl from '@doedja/scenecut-core/wasm-binary?url';
```

--------------------------------

### Export scene changes to FCPXML format

Source: https://github.com/doedja/scenecut/blob/main/README.md

Generate a Final Cut Pro X timeline with scene markers by exporting scene changes using the fcpxml format.

```bash
scenecut video.mkv -f fcpxml
```

--------------------------------

### Import WASM Glue Code (Vite)

Source: https://github.com/doedja/scenecut/blob/main/packages/core/README.md

Import the ES-module glue code for the WASM binary using Vite's URL import syntax. This is a shortcut for importing the compiled WASM module.

```typescript
import glueUrl from '@doedja/scenecut-core/wasm?url';        // via Vite
```

--------------------------------

### Build Core TypeScript Module

Source: https://github.com/doedja/scenecut/blob/main/README.md

Builds the core TypeScript components of the project.

```bash
bun run build:core:ts
```

--------------------------------

### Default Scene Cut Detection

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Run scenecut with default settings to detect scene changes and output Aegisub keyframes.

```bash
scenecut input.mkv
```

--------------------------------

### Export scene changes to JSON format

Source: https://github.com/doedja/scenecut/blob/main/README.md

Save the detailed scene change detection results, including metadata and statistics, in a JSON file.

```bash
scenecut video.mp4 -f json -o scenes.json
```

--------------------------------

### Advanced Scene Cut Detection Options

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Configure scene cut detection with higher sensitivity and verbosity, or set a timeout to abort processing.

```bash
scenecut anime.mkv -s high -v          # higher sensitivity, verbose
```

```bash
scenecut long.mkv -t 120               # abort after 2 minutes
```

```bash
scenecut video.mp4 --thumbnails ./thumbs
```

--------------------------------

### Export scene changes to CSV format

Source: https://github.com/doedja/scenecut/blob/main/README.md

Generate a generic CSV file containing scene change information such as frame number, timestamp, timecode, confidence, and duration.

```bash
scenecut video.mp4 -f csv  -o scenes.csv
```

--------------------------------

### Low-level Source Imports

Source: https://github.com/doedja/scenecut/blob/main/packages/web/README.md

Importing specific frame source classes for direct control over video frame extraction. These classes implement the FrameSource interface from @doedja/scenecut-core.

```typescript
import {
  WebCodecsSource,   // MP4 / MOV
  MkvSource,         // MKV / WebM
  VideoElementSource // fallback
} from '@doedja/scenecut-web';
```

--------------------------------

### Parallel Scene Cut Detection with Worker Pool

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Utilize a worker pool for parallel analysis to speed up scene detection. The number of workers can be specified or set to auto.

```bash
scenecut video.mkv -w true             # auto (cpus - 1, capped at 8)
```

```bash
scenecut video.mkv -w 4                # pin to 4 workers
```

--------------------------------

### Export scene changes to timecode format

Source: https://github.com/doedja/scenecut/blob/main/README.md

Output scene changes as a plain list of timecodes in HH:MM:SS.mmm format, one per line.

```bash
scenecut video.mp4 -f timecode
```

--------------------------------

### Scene Cut Detection for Technical Formats

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Output scene cut information in technical formats like JSON, CSV, or timecode.

```bash
scenecut video.mp4 -f json -o scenes.json
```

```bash
scenecut video.mp4 -f csv
```

```bash
scenecut video.mp4 -f timecode
```

--------------------------------

### Export Scene Detection Results

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Utilize exporter functions like `formatEdl`, `formatFcpxml`, and `formatPremiereMarkers` to convert the detection results into standard editing formats.

```javascript
const { formatEdl, formatFcpxml, formatPremiereMarkers } = require('@doedja/scenecut');
const edl = formatEdl(result, 'my-clip');
```

--------------------------------

### Set processing timeout

Source: https://github.com/doedja/scenecut/blob/main/README.md

Limit the processing time for a video file by specifying a timeout in seconds. The process will abort if it exceeds this duration.

```bash
scenecut long-movie.mkv -t 120
```

--------------------------------

### Scene Cut Detection for NLE Exports

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Generate scene cut data in formats compatible with Non-Linear Editors (NLEs) like Premiere, Resolve, and Final Cut Pro X.

```bash
scenecut video.mkv -f edl              # CMX3600 — Premiere, Resolve, Avid
```

```bash
scenecut video.mkv -f fcpxml           # Final Cut Pro X timeline with markers
```

```bash
scenecut video.mkv -f premiere         # Premiere Pro Import Markers CSV
```

--------------------------------

### Export scene changes to EDL format

Source: https://github.com/doedja/scenecut/blob/main/README.md

Use scenecut to detect scene changes and export them in CMX3600 EDL format, compatible with professional NLE software like Premiere, Resolve, and Avid.

```bash
scenecut video.mkv -f edl
```

--------------------------------

### Detect Scene Changes in Node.js

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Use the `detectSceneChanges` function to analyze a video file and identify scene transitions. Configure sensitivity, enable workers for performance, and provide callbacks for progress and scene detection events.

```javascript
const { detectSceneChanges } = require('@doedja/scenecut');

const result = await detectSceneChanges('input.mp4', {
  sensitivity: 'low',
  workers: true,
  onProgress: (p) => console.log(`${p.percent}% — ${p.fps?.toFixed(1)} fps`),
  onScene:    (s) => console.log(`cut @ ${s.timecode} conf=${s.confidence?.toFixed(2)}`)
});

console.log(`${result.scenes.length} scenes`);
```

--------------------------------

### Extract Scene Images

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Use the `extractSceneImages` function to generate thumbnail images for detected scenes. Specify output directory, image format, and quality settings.

```javascript
const { extractSceneImages } = require('@doedja/scenecut');

await extractSceneImages('input.mp4',
  { sensitivity: 'low' },
  { outputDir: './thumbs', format: 'jpg', quality: 85 }
);
```

--------------------------------

### Export scene changes for Premiere Pro

Source: https://github.com/doedja/scenecut/blob/main/README.md

Export scene changes in a CSV format suitable for importing as markers in Adobe Premiere Pro using the 'File → Import Markers' option.

```bash
scenecut video.mkv -f premiere
```

--------------------------------

### Cancel Scene Detection with AbortSignal

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Implement cancellation for long-running scene detection tasks by passing an `AbortSignal` to the `detectSceneChanges` function. This allows for graceful termination of the process.

```javascript
const ctrl = new AbortController();
setTimeout(() => ctrl.abort(), 60_000);
await detectSceneChanges('input.mp4', { signal: ctrl.signal });
```

--------------------------------

### Configure parallel worker pool

Source: https://github.com/doedja/scenecut/blob/main/README.md

Control the number of parallel workers used for scene detection. Set to 'true' for automatic detection (CPU cores - 1, capped at 8), a specific number, or 'off' to disable.

```bash
scenecut video.mkv -w true
```

```bash
scenecut video.mkv -w 4
```

--------------------------------

### Exporter Imports

Source: https://github.com/doedja/scenecut/blob/main/packages/web/README.md

Importing various formatting functions to export detection results into different standard formats like EDL, FCPXML, and Premiere markers. These are re-exported from the core library.

```typescript
import { formatEdl, formatFcpxml, formatPremiereMarkers } from '@doedja/scenecut-web';
const edl = formatEdl(result, 'my-clip');
```

--------------------------------

### Scene Detection Result Interface

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Describes the structure of the `DetectionResult` object returned by `detectSceneChanges`, including details about detected scenes, video metadata, and processing statistics.

```typescript
interface DetectionResult {
  scenes: Array<{ 
    frameNumber: number;
    timestamp: number;   // seconds
    timecode: string;    // HH:MM:SS.mmm
    confidence: number;  // 0–1 (sigmoid-calibrated)
    duration: number;
    frameCount: number;
  }>;
  metadata: { totalFrames, duration, fps, resolution, codec?, pixelFormat?, bitrate? };
  stats:    { processingTime, framesPerSecond };
}
```

--------------------------------

### Scene Detection Result Interface

Source: https://github.com/doedja/scenecut/blob/main/README.md

Describes the structure of the `DetectionResult` object returned by `detectSceneChanges`, including details about detected scenes and video metadata.

```typescript
interface DetectionResult {
  scenes: Array<{ 
    frameNumber: number;
    timestamp: number;   // seconds
    timecode: string;    // HH:MM:SS.mmm
    confidence: number;  // 0–1
    duration: number;
    frameCount: number;
  }>;
  metadata: { totalFrames, duration, fps, resolution, codec?, pixelFormat?, bitrate? };
  stats:    { processingTime, framesPerSecond };
}
```

--------------------------------

### Node Detection Options Interface

Source: https://github.com/doedja/scenecut/blob/main/packages/node/README.md

Defines the available options for the `detectSceneChanges` function, including sensitivity, search range, worker configuration, and event callbacks.

```typescript
interface NodeDetectionOptions {
  sensitivity?: 'low' | 'medium' | 'high';             // default: 'low'
  searchRange?: 'auto' | 'small' | 'medium' | 'large'; // default: 'auto'
  workers?: number | boolean;                          // default: off
  onProgress?: (p: Progress) => void;
  onScene?: (s: SceneInfo) => void;
  signal?: AbortSignal;
}
```

--------------------------------

### Reveal Body on Load

Source: https://github.com/doedja/scenecut/blob/main/apps/web/index.html

This JavaScript reveals the body element once the CSS and fonts have been painted at least once, ensuring a smooth visual experience.

```javascript
document.addEventListener('DOMContentLoaded', () => {
  requestAnimationFrame(() => document.body.classList.add('is-ready'));
});
```

--------------------------------

### Critical CSS for scenecut

Source: https://github.com/doedja/scenecut/blob/main/apps/web/index.html

This CSS is critical for preventing Flash of Unstyled Content (FOUC) while the main stylesheet loads. It ensures a smooth transition by matching the main stylesheet's variables.

```css
html, body { margin: 0; background: #0a0a0b; color: #e8e6df; font-family: 'IBM Plex Sans', ui-sans-serif, system-ui, -apple-system, sans-serif; -webkit-font-smoothing: antialiased; color-scheme: dark; }
[hidden] { display: none !important; }
body { opacity: 0; transition: opacity 120ms ease-out; }
body.is-ready { opacity: 1; }
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.