[FFmpeg-devel] [PATCH v2] avformat/matroska: Write WebVTT subtitles according to MKV specs

[Andreas Rheinhardt]

Gwyneth Morgan:
> When writing WebMs, FFmpeg muxes WebVTT subtitles with the D_WEBVTT/*
> codec tags from the WebM specs [1]. However, it does the same when
> muxing MKV files, and the Matroska specifications instead use
> S_TEXT/WEBVTT tags for WebVTT subtitles [2], which FFmpeg currently
> doesn't understand. Support reading MKVs using either tag, write regular
> MKVs with S_TEXT/WEBVTT, and write WebMs with the D_WEBVTT/* tags we
> already use.
> 
> [1]: https://www.webmproject.org/docs/container/
> [2]: https://matroska.org/technical/codec_specs.html#s_textwebvtt
> 
> Signed-off-by: Gwyneth Morgan <gwymor at tilde.club>
> Fixes: https://trac.ffmpeg.org/ticket/5641
> ---
> v2: Rebase as later changes in ffmpeg.git conflicted with this patch.
> 
>  libavformat/matroska.c | 11 ++++++-----
>  1 file changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/libavformat/matroska.c b/libavformat/matroska.c
> index 79b2d09..d9e5ff9 100644
> --- a/libavformat/matroska.c
> +++ b/libavformat/matroska.c
> @@ -60,16 +60,12 @@ const CodecTags ff_mkv_codec_tags[]={
>      {"A_VORBIS"         , AV_CODEC_ID_VORBIS},
>      {"A_WAVPACK4"       , AV_CODEC_ID_WAVPACK},
>  
> -    {"D_WEBVTT/SUBTITLES"   , AV_CODEC_ID_WEBVTT},
> -    {"D_WEBVTT/CAPTIONS"    , AV_CODEC_ID_WEBVTT},
> -    {"D_WEBVTT/DESCRIPTIONS", AV_CODEC_ID_WEBVTT},
> -    {"D_WEBVTT/METADATA"    , AV_CODEC_ID_WEBVTT},
> -
>      {"S_TEXT/UTF8"      , AV_CODEC_ID_SUBRIP},
>      {"S_TEXT/UTF8"      , AV_CODEC_ID_TEXT},
>      {"S_TEXT/ASCII"     , AV_CODEC_ID_TEXT},
>      {"S_TEXT/ASS"       , AV_CODEC_ID_ASS},
>      {"S_TEXT/SSA"       , AV_CODEC_ID_ASS},
> +    {"S_TEXT/WEBVTT"    , AV_CODEC_ID_WEBVTT},
>      {"S_ASS"            , AV_CODEC_ID_ASS},
>      {"S_SSA"            , AV_CODEC_ID_ASS},
>      {"S_VOBSUB"         , AV_CODEC_ID_DVD_SUBTITLE},
> @@ -78,6 +74,11 @@ const CodecTags ff_mkv_codec_tags[]={
>      {"S_HDMV/TEXTST"    , AV_CODEC_ID_HDMV_TEXT_SUBTITLE},
>      {"S_ARIBSUB"        , AV_CODEC_ID_ARIB_CAPTION},
>  
> +    {"D_WEBVTT/SUBTITLES"   , AV_CODEC_ID_WEBVTT},
> +    {"D_WEBVTT/CAPTIONS"    , AV_CODEC_ID_WEBVTT},
> +    {"D_WEBVTT/DESCRIPTIONS", AV_CODEC_ID_WEBVTT},
> +    {"D_WEBVTT/METADATA"    , AV_CODEC_ID_WEBVTT},
> +
>      {"V_AV1"            , AV_CODEC_ID_AV1},
>      {"V_AVS2"           , AV_CODEC_ID_AVS2},
>      {"V_AVS3"           , AV_CODEC_ID_AVS3},

The reason we write it the way we do is that webvtt is muxed differently
in Matroska than WebM. This needs to be fixed, too, before S_TEXT/WEBVTT
can be used for Matroska.

- Andreas