Subtext

Subtext is a subtitle renderer that uses libass and ffmpeg.

.. function:: TextFile(clip clip, string file[, string charset="UTF-8", float scale=1, int debuglevel=0, string fontdir="", float linespacing=0, int[] margins=[0, 0, 0, 0], float sar=0, string style="", bint blend=True, int matrix, string matrix_s, int transfer, string transfer_s, int primaries, string primaries_s]) :module: sub

TextFile renders text subtitles. Supported formats include ASS, JACOsub, MicroDVD, SAMI, SRT, WebVTT, and some other obscure ones.

TextFile has two modes of operation. With blend=True (the default), it returns clip with the subtitles burned in. With blend=False, it returns an RGB24 clip containing the rendered subtitles, with a Gray8 frame attached to each frame in the _Alpha frame property. These Gray8 frames can be extracted using std.PropToClip.

Parameters: clip Input clip. It must have a constant format, constant dimensions, and a constant frame rate.

  file
     Subtitle file to be rendered.

  charset
     Character set of the subtitle, in iconv format.

  scale
     Font scale.

  debuglevel
     Debug level. Increase to make libass more chatty.
     See `ass_utils.h <https://github.com/libass/libass/blob/master/libass/ass_utils.h>`_
     in libass for the list of meaningful values.

  fontdir
     Directory with additional fonts.

  linespacing
     Space between lines, in pixels.

  margins
     Additional margins, in pixels. Negative values are not
     allowed. The order is top, bottom, left, right.

  sar
     Sample aspect ratio, also known as the pixel aspect ratio: the
     ratio of the width of a pixel to its height. 1 means square
     pixels.

     By default this follows the ``_SARNum`` and ``_SARDen`` frame
     properties of *clip*, individually for every frame, so
     anamorphic video gets subtitles that look right once the video
     is displayed at its intended shape. Frames without those
     properties are treated as having square pixels.

     Setting it uses the given ratio for every frame instead. That
     also means no frame of *clip* has to be looked at, which is
     worth doing when blend=False and producing the frames of *clip*
     is expensive.

  style
     Custom ASS style for subtitle formats other than ASS. If empty
     (the default), libavcodec's default style is used. This
     parameter has no effect on ASS subtitles.

  blend
     If True, the subtitles will be blended into *clip*. Otherwise,
     the bitmaps will be returned untouched.

  matrix

  matrix_s

  transfer

  transfer_s

  primaries

  primaries_s
     Overrides for the colorimetry of *clip*. Only used when
     blend=True, and only needed when *clip* is tagged wrongly or
     not at all.

     By default the subtitles are converted to match *clip*, frame
     by frame, following its ``_Matrix``, ``_Transfer``,
     ``_Primaries``, ``_ColorRange`` and ``_ChromaLocation`` frame
     properties. The subtitles are treated as BT.709 to start with,
     so a frame tagged as, say, PQ and BT.2020 gets subtitles
     converted into that colour space instead of blended in as if
     they were already in it. A frame that says nothing usable is
     handled the way it always was: the BT.709 matrix and no colour
     space conversion. The colour range is always taken from *clip*
     and can't be overridden, because the two can only be blended
     when they agree on it.

     These are worth reaching for in two cases:

     Untagged clips. Nothing can be adapted to, so the subtitles
     fall back to BT.709. Standard definition material is the usual
     victim, and ``matrix_s="470bg"`` is the usual answer.
     Tagging the clip with std.SetFrameProps instead is generally
     better, since that fixes everything downstream and not just
     the subtitles.

     Turning the colour space conversion off. Passing
     ``transfer=1, primaries=1`` converts the subtitles from BT.709
     to BT.709, which does nothing, so they end up blended in
     unconverted the way they were before this was done per frame.
     Both are needed, since either one on its own still leaves the
     other being adapted to.

.. function:: Subtitle(clip clip, string text[, int start=0, int end=clip.numFrames, int debuglevel=0, string fontdir="", float linespacing=0, int[] margins=[0, 0, 0, 0], float sar=0, string style="sans-serif,20,&H00FFFFFF,&H000000FF,&H00000000,&H00000000,0,0,0,0,100,100,0,0,1,2,0,7,10,10,10,1", bint blend=True, int matrix, string matrix_s, int transfer, string transfer_s, int primaries, string primaries_s]) :module: sub

Instead of rendering a subtitle file, Subtitle renders the string text. Otherwise it works the same as TextFile.

Parameters: text String to be rendered. This can include ASS tags to enable rich text and animation.

  style
     Custom ASS style to be used.
  
  start, end
     Subtitle will be shown from *start* up until *end*. By default this will be for all frames in *clip*.

The other parameters have the same meanings as with TextFile.

.. function:: ImageFile(clip clip, string file[, int id=-1, int[] palette, bint gray=False, bint info=False, bint flatten=False, bint blend=True, int matrix, string matrix_s, int transfer, string transfer_s, int primaries, string primaries_s]) :module: sub

ImageFile renders image-based subtitles such as VOBSUB and PGS.

Parameters: clip If blend is True, the subtitles will be burned into this clip, Otherwise, only the frame rate and number of frames will be obtained from this clip.

     It must have a constant format, constant dimensions, and a
     constant frame rate, even when *blend* is False.

  *file*
     Name of the subtitle file. For VOBSUB, it must the name of the
     idx file. The corresponding sub file must be in the same
     folder, and it must have the same name.

  *id*
     Id of the subtitle track to render. There may be several
     subtitle tracks in the same file. If this is -1, the first
     supported subtitle track will be rendered. Use info=True to
     see a list of all subtitle tracks, including their ids.

     Default: -1.

  *palette*
     Custom palette. This is an array of at most 256 integers. Each
     element's least significant four bytes must contain the values
     for alpha, red, green, and blue, in that order, from most
     significant to least.

     Additionally, the special value 2**42 means that the
     corresponding element of the original palette is used. This
     way it is possible to override only the third element, without
     overriding the first and second ones, for example.

     An alpha value of 255 means the colour will be completely
     opaque, and a value of 0 means the colour will be completely
     transparent.

  *gray*
     If True, the subtitles will be turned gray.

     Default: False.

  *info*
     If this is True, a list of all supported subtitle tracks found
     in the file will be printed on each frame of the output. The
     information printed about each track includes the id, the
     language (if known), the resolution, and the format.

     Default: False.

  *flatten*
     If this is True, ImageFile will output a clip with exactly as
     many frames as there are pictures in the subtitle file.

     If this is True, *blend* has no effect (no automatic blending).

     Default: False.

The other parameters have the same meanings as with TextFile.

Example with manual blending::

sub = core.sub.TextFile(clip=YUV420P10_video, file="asdf.ass", blend=False) mask = core.std.PropToClip(clip=sub, prop='_Alpha')

gray10 = core.query_video_format( mask.format.color_family, YUV420P10_video.format.sample_type, YUV420P10_video.format.bits_per_sample, mask.format.subsampling_w, mask.format.subsampling_h )

sub = core.resize.Bicubic(clip=sub, format=YUV420P10_video.format.id, matrix_s="470bg") mask = core.resize.Bicubic(clip=mask, format=gray10.id)

hardsubbed_video = core.std.MaskedMerge(clipa=YUV420P10_video, clipb=sub, mask=mask)

Example with automatic blending (will use BT709 matrix)::

hardsubbed_video = core.sub.TextFile(clip=YUV420P10_video, file="asdf.ass")

Example with a custom palette and automatic blending::

def rgba(r, g, b, a=255): if r < 0 or r > 255 or g < 0 or g > 255 or b < 0 or b > 255 or a < 0 or a > 255: raise vs.Error("Colours must be in the range [0, 255].")

   return (a << 24) + (r << 16) + (g << 8) + b

unused = 1 << 42

src = core.ffms2.Source("video.mp4")

Override only the third element of the palette. Set it to some kind of green.

ret = core.sub.ImageFile(src, "subtitles.sup", palette=[unused, unused, rgba(0, 192, 128)])