Here is a table with the syntactical differences of different existing subtitle formats for which Chris Chiu wrote conversion scripts to CMML.

Structural Syntax Information:

• File Extensions:
  • They are not standardised for all subtitle formats. Hence, the syntax
    and structure of the subtitle is important to verify to determine the
    subtitle type, rather than from the file extension alone.
  • As an example, MicroDVD, MPSub, SubViewer and VOBSub all use the same
    extension.
• Subtitle Type:
  • Subtitle type refers to the formatting of the subtitle format that is
    employed.
  • Text-based subtitles use a simplified ASCII text notation to define the
    start and end times/frames of the subtitles, and to define the text
    positioning and styles if applicable.
  • HTML-based subtitles use HTML style tags to define timings and text
    formatting. However, these formats do not strictly adhere to XML standards
    as attributes may not be in quotation marks, or that closing tags can be
    missing.
  • XML-based subtitles adhere to XML Schema standards, with DTD's to define
    the structure and syntax of the subtitles. Parsing is being implemented
    using the XML Parser Libraries for Python
  • Image-based subtitles use raster images (compressed JPEG images
    typically) to display subtitles. These kind of subtitle formats typically
    encapsulate the subtitle images in one layer, while the synchronisation of
    the subtitle images are stored in a separate layer. Hence, the extraction of
    image-based subtitles, depending on the format, can result in more than one
    file.
• Character Encoding:
  • The character encoding is important as it means that the parsed CMML
    script can be strictly validated.
  • While some of the subtitle formats define Unicode as the character
    encoding type, older formats tend to have no encoding standard whatsoever
    (and so by assumption, plain old ASCII).
  • The scripts do not currently convert non-unicode characters to their
    unicode equivalents.
• Line Breaks:
  • It is important to determine whether line breaks are used for text, HTML
    or XML subtitle formats. This is because CMML does not currently support the
    use of line breaks as of present. When parsing a subtitle line break, the
    two methods with parsing line breaks is to remove all line breaks and
    concatenate the lines together, or recreate the line break with indentation.
    The second method would be used to make the CMML file more readable, but
    transcoding the Ogg Theora file and CMML file to Annodex may result in the
    line breaks being ignored.
• Text Style Type:
  • Some formats support the use of text styles to allow the positioning and
    formatting of subtitles shown on the screen. The degree of text manipulation
    will vary depending on the format if it is specifically defined by the
    original subtitle creators, but the HTML subtitle formats typically employ
    standards-based text formatting such as SMIL and CSS. XML subtitle formats
    can either specify the styling from the DTD, or use standards-based
    formatting.
  • Text styles not in CSS are not supported by the scripts at the moment. A
    future functionality would be to convert these styles into CSS, but it must
    be noted that the variety of text styles used by some subtitle formats may
    be poorly documented and/or difficult to obtain.
• Metadata Info:
  • Some of the subtitle formats support metadata information to detail the
    video title, the names and authors of the subtitles, as well as a variety of
    attributes to store extraneous information. Some of this data is parsed by
    the scripts into CMML, while others have been omitted. Some subtitle formats
    include technical data including the file hash and size, as well as
    copyright information.
• Timings:
  • Timings are of particular importance to subtitles for the most obvious
    reason: to ensure the correct synchronisation of subtitles to the video and
    audio playback.
  • As CMML timings are based on an elapsed time format (Hours, Minutes,
    Seconds), the subtitles that employ the same timing method is the easiest to
    convert into CMML without problems.
  • Sequential timing formats (which is used by MPSub) is different in that
    the subtitles are timed depending on when the last subtitle was displayed.
    The addition of these times results converts these timings to an elapsed
    time format.
  • The older formats use the video/audio frame to synchronise the subtitles.
    The main issue that arises is that we need to know the frames per second
    (fps) value of the media before the timings can be converted into an elapsed
    time format.
• Timing Precision:
  • The timing precision of most subtitles is 10 milliseconds (1x10^-2 sec),
    but newer subtitle formats employ greater precisions up to 1 millisecond
    (1x10^-3 sec).
  • Note for Frame timing subtitles converted into elapsed times:
    • They are converted to a level of 1 millisecond precision due to the
      potential of rounding up/down errors from fractional conversion between
      timing values.