[Ffmpeg-devel] [REQUEST] Unicode in the -title argument for PSP

[Chris Dolan]

== My problem ==
I'm trying to create movies with non-ASCII titles for the Sony PSP.   
The movies play great, but their titles contain '?' characters  
instead of the proper Unicode characters in the PSP's video listing.   
I'm using a build from today's SVN head (revision 5456).

== Details ==
The Sony PSP reads the MPEG4 "moov.uuid.MTDT" atom to get the title  
of the video file for its onboard directory listing.  This field gets  
populated via the ffmpeg cmdline flag "-title <string>".  Looking in  
libavcodec/movenc.c, I see that this string is downsampled to ASCII  
and then upgraded to UTF-16BE.  This is done by changing any values  
outside 0x20 and 0x7f to a question mark ('?') and then padding to 16  
bits via the ascii_to_wc() function.  It looks like this was done as  
a simple expedient since the PSP expects a UTF-16 string (I believe).

The command line I tried is the following.  It contains the greek  
letter lambda encoded via UTF-8.

   ffmpeg -v 1 -y -i mjpegA.mov -title 'The ?-Orionis Star Forming  
Region' -s 368x208 -r 29.97 -b 550 -maxrate 700 -bufsize 700 -acodec  
aac -ac 2 -ar 24000 -ab 48 -f psp M4V88803.MP4

== My request ==
I'd like a way to be able to provide a Unicode string for the title  
value.  I'm fine with any way to provide this data.  Some options off  
the top of my head:
  * UTF-16BE read from a file (note: would you have to strip the BOM?)
  * UTF-8 on the cmdline converted to UTF-16 internally
  * \unnnn escapes in cmdline string converter to UTF-16 internally

The latter is the easiest to code, but could break compatibility with  
users that already have backslashes in their titles.  I would be  
willing to implement this if others approve of the idea.

Thanks!
Chris
--
Chris Dolan, Software Developer, http://www.chrisdolan.net/
Public key: http://www.chrisdolan.net/public.key
vCard: http://www.chrisdolan.net/ChrisDolan.vcf