[FFmpeg-devel] Matroska subtitle decoding: CODEC_ID_TEXT vs. CODEC_ID_SRT

[Martin Lambers]

Hello Reimar,

thanks for having a look at this.

On 24/04/11 17:16, Reimar Döffinger wrote:
> On Sun, Apr 24, 2011 at 02:54:09PM +0200, Martin Lambers wrote:
>> According to this site
>> <http://matroska.org/technical/specs/codecid/index.html>, "S_TEXT/UTF8"
>> are SRT subtitles.
> 
> What they are writing is complete nonsense.
> Quote:
> "When placing SRT in Matroska, part 3 is converted to UTF-8 and placed in the data portion of the Block. Part 2 is used to set the timecode of the Block, and BlockDuration element. Nothing else is used."
> Or in other words they don't write any SRT data into the file at all,
> but plain text. From what it looks like whoever wrote it was unable
> of distinguishing between the mkvmerge tool and the MKV format.
> I see really no relation at all between SRT and what MKV stores
> there, except that the conversion is really easy.

I agree that the wording of the text is confused. My interpretation is
that when they say "plain text" or "UTF-8 text" they really mean SRT
text with the UTF-8 character encoding (this would apply to part 3).

I think the intention here is that SRT subtitles can be stored in the
Matroska container, namely as "S_TEXT/UTF8".

>> Yet in libavformat/matroska.c, there are two entries
>> for "S_TEXT/UTF8":
>>
>> ...
>> {"S_TEXT/UTF8"      , CODEC_ID_TEXT},
>> {"S_TEXT/UTF8"      , CODEC_ID_SRT},
>> ...
>>
>> This has the result that SRT subtitles are handled with CODEC_ID_TEXT
>> during decoding since this is the first entry found. Therefore, they do
>> not get translated to ASS as would happen with SRT subtitles from other
>> sources.
> 
> CODEC_ID_TEXT is correct by what I can tell, it might help if
> you actually told us what specifically the issue is you have.

The issue is that if SRT subtitles are part of a .mkv file, they are
stored as "S_TEXT/UTF8", and FFmpeg will decode them as plain text
subtitles, and thus a video player based on FFmpeg will not properly
interpret SRT tags such as <i>, <b> etc. This was reported by a user of
the Bino video player.

>> Is there a reason to keep this behaviour? Can the first entry be removed?
> 
> Removing it would probably break muxing in text subtitles.
> You could just change their order around, but IMO that's still wrong.

>From my understanding of the text, I think it would be correct to change
the order. Otherwise, there would be no way to decode SRT subtitles
stored in a Matroska container.

And this change should do no harm, either, since interpreting plain text
as SRT has no effects unless the plain text contains SRT tags.

Martin