Javicle - a JSON Video Composition Language

JVCL (pronounced “Javicle”) is a JSON DSL for audio/video transformations.

Under the hood, it’s all shell commands: ffmpeg, mediainfo, sox, and so on.

JVCL provides higher-level semantics for working with these lower level tools.

A Quick Example

Say you want to split a portion of a video into ten-second chunks. With ffmpeg and bash, you might do something like this:

    INCR=10
    for ((i=10;i<130;i=(i+INCR))); do
      ffmpeg -i /tmp/my/source.mp4 -ss ${i} -t $((i+INCR)) /tmp/my/slice_${i}_$((i+INCR)).mp4
    done

With JVCL, you’d create this spec:

{
  "assets": [ {"name": "src", "path": "/tmp/my/source.mp4"} ],
  "operations": [{
    "operation": "split",
    "creates": "src_splits",
    "perform": {
      "split": "src",
      "interval": "10s",
      "start": "10s",
      "end": "130s"
    }
  }]
}

Yes, the JVCL is longer, but I think many would agree it is easier to read and maintain.

As the number of input assets and operations grows, hand-crafted shell scripts with magical ffmpeg incantations become ever more inscrutable.

JVCL is designed for readability and maintainability. JVCL will continue to evolve towards greater coverage of the full capabilities of ffmpeg. We also plan to introduce “function” concepts to create reusable compound operations, further increasing reusability and lowering long-term maintenance.

Who is JVCL not for?

If you like GUIs, JVCL is probably not for you.

JVCL is not a replacement for Final Cut Pro or even iMovie.

Who is JVCL for?

If you like CLIs, JVCL might be for you.

You might enjoy JVCL if your video composition needs are relatively simple or if you enjoy capturing repeatable processes in source control.

Concepts

In JVCL there are two main concepts: assets and operations.

Assets

Assets are the inputs: generally image, audio and video files. Assets have a name and a path.

The path can be a file or a URL.

Operations

Operations are transformations to perform on the inputs.

An operation can produce one or more new assets, which can then be referenced in later operations.

The operations that JVCL either supports or intends to support are:

split

Split an audio/video asset into multiple assets

concat

Concatenate audio/video assets together into one asset

trim

Trim audio/video; crop a section of an asset, becomes a new asset

overlay

Overlay one audio or video file onto another

ken-burns

For transforming still images into video via a fade-pan (aka Ken Burns) effect

letterbox

Transform a video in one size to another size using black letterboxes on the sides or top/bottom. Handy for embedding mobile videos into other screen formats

split-silence

Split an audio file according to silence

Complex Example

Here is a complex example using multiple assets and operations:

{
  "assets": [
    {"name": "vid1", "path": "/tmp/path/to/video1.mp4"},
    {"name": "vid2", "path": "/tmp/path/to/video2.mp4"}
  ],
  "operations": [
    {
      "operation": "split",            // name of the operation
      "creates": "vid1_split_%",       // assets it creates, the '%' will be replaced with a counter
      "perform": {
        "split": "vid1",               // split this source asset
        "interval": "10s"              // split every ten seconds
      }
    },
    {
      "operation": "concat",           // name of the operation
      "creates": "recombined_vid1",    // assets it creates, the '%' will be replaced with a counter
      "perform": {
        "concat": ["vid1_split"]       // recombine all split assets
      }
    },
    {
      "operation": "concat",           // name of the operation
      "creates": "combined_vid",       // asset it creates, can be referenced later
      "perform": {
        "concat": ["vid1", "vid2"]     // operation-specific: this says, concatenate these named assets
      }
    },
    {
      "operation": "concat",           // name of the operation
      "creates": "combined_vid",       // the asset it creates, can be referenced later
      "perform": {
        "concat": ["vid1", "vid2"]     // operation-specific: this says, concatenate these named assets
      }
    },
    {
      "operation": "overlay",          // name of the operation
      "creates": "overlay1",           // asset it creates
      "perform": {
        "source": "combined_vid1",     // main video asset
        "overlay": "vid2",             // overlay this video on the main video

        "offset": "30",                // when (on the main video timeline) to begin showing the overlay. default is 0 (beginning)
        "overlayStart": "0",           // when (on the overlay video timeline) to begin playback on the overlay. default is 0 (beginning)
        "overlayEnd": "0",             // when (on the overlay video timeline) to end playback on the overlay. default is to play the whole overlay

        "width": "overlay.width / 2",  // how wide the overlay will be, in pixels. default is the full overlay width, or maintain aspect ratio if height was set
        "height": "",                  // how tall the overlay will be, in pixels. default is the full overlay height, or maintain aspect ratio if width was set

        "x": "source.width/2",         // horizontal overlay position on main video. default is 0
        "y": "source.height/2",        // vertical overlay position on main video. default is 0

        "outputWidth": "1920",         // output width in pixels. default is source width
        "outputHeight": "1024"         // output height in pixes. default is source height
      }
    }
  ]
}

What’s up with the name?

I dunno, a cross between a javelin and an icicle? does that have any positive connotations? ok then...