Dialogger

Dialogger is an audio/video editor that allows you to navigate and edit recordings using a text-based interface.

What's included

Playback and navigation using transcript
Transcript editing
Export of edit decision list (EDL)
User accounts
Asset management

What's not included

The following features must be added manually for Dialogger to operate fully. Instructions and examples are provided.

Speech-to-text
Preview file generator
File export

Conceptual flow diagram (excluded features shown in red)

Technology stack

Front-end

HTML/CSS/JS
Text editing: CKEditor (LGPL)
Media playback: HTML5 Video Compositor (Apache-2.0)
UI framework: Semantic UI (MIT)
MVC framework: Backbone (MIT)
File upload: Dropzone (MIT)

Back-end

Node.js / Express
Database: MongoDB (Apache-2.0)
Authentication: Passport.js (MIT)
Media info: Mimovie (MIT)
Logging: Bunyan (MIT)

Installation

Using Docker (recommended)

git clone https://github.com/bbc/dialogger.git && cd dialogger
docker-compose build
docker-compose up

Navigate to http://localhost:8080 and log in with username user and password password.

Ubuntu/Debian

git clone https://github.com/bbc/dialogger.git && cd dialogger
curl -sL https://deb.nodesource.com/setup_6.x | sudo -E bash -
sudo -E apt-get install -y nodejs mediainfo mongodb
sudo -E npm install -g gulp bower bunyan
npm install
npm run build
(cd data && ./import.sh)

In config/consts.js, set the following parameters:

consts.port
consts.db.url
consts.cookie.serverDomain
consts.cookie.serverPath
consts.files.root (ensure write permissions are set)

Run Dialogger using npm start, then log in with username user and password password.

Configuration

Dialogger does not include the following key functionality, so you must add this in manually. Instructions on how to add this are provided below.

Speech-to-text
Preview file generator
File export

1. Speech-to-text

Dialogger does not come with a speech-to-text system, so you will need to add some code to helpers/stt.js that accepts a path to an audio/video file and returns the transcript and segmentation data. Examples of the data formats are shown below, and an example can be found in helpers/stt-example.js.

Transcript format

{
  words: [
    {
      word: "hello",
      punct: "Hello",
      start: 0.05,
      end: 0.78,
      confidence: 0.45
    },
    {
      word: "world",
      punct: "world.",
      start: 1.13,
      end: 1.45,
      confidence: 0.9
    }
  ]
}

Segments format

{
  segments: [
    {
      start: 0.05,
      duration: 2.34,
      speaker: {
        @id: "Bob",
        gender: "M"
      }
    },
    {
      start: 2.34,
      duration: 4.2,
      speaker: {
      @id: "Alice",
        gender: "F"
      }
    }
  ]
}

2. Preview file generator

Preview files are low-bitrate versions of media files which are used for playback in the browser interface. To configure preview file generation, you will need to add some code to helpers/previewfile.js. The function should receive options in the following format, create a preview file and run the callback function.

Options format

{
  inputPath: "/path/to/input/file",
  outputPath: "/path/to/preview/version",
  format: "audio",  // can be audio or video
  audio: {
    acodec: "aac",
    ab: "128k"
  }    
}

3. File export

File export allows users to download an edited version of their media. To configure file export, you will need to add some code to helpers/fileexport.js. The function should receive options in the following format and return the path of the edited file. In essence, what you want to do is to take the file path (asset.path) and the list of edits (edl), produce an edited version of the file, then return the path.

Options format

{
  // Information about the original file/asset
  asset: {
    description: "Asset description",
    filename: "AssetFilename.wav",
    path: "/path/to/original/file",
    audio: {
      channels: 2,
      sampleRate: "48000"
    }
  },
  
  // An array of in- and out-points, in seconds
  edl: [
    ["78.38","102.89"],
    ["128.3","135.17"]
  ],
  
  // User-provided options from the exportForm
  //   in public/js/editor.html
  settings: {
    audio: {
      ab: "",
      acodec: "pcm_s16le"
    },
    edlformat: "dira",
    exportUnderlined:"true",
    format: "audio",
    id: "",
    include: "true",
    name: "test.wav",
    video: {
      height: "",
      vb: "",
      f: "mp4",
      acodec: "aac",
      ab: "",
      vcodec: "libx264",
      width: ""
    }
  }
}

Issues/development

Please report any problems or make feature requests by raising an issue. Pull requests are also welcome.

Authors

Chris Baume, BBC Research and Development