[FFmpeg-user] how ffmpeg computes(find)canvas_size for subtitles

[cavelier]

Hi.
I record broadcast from French TNT (1920:1080i and subtitles dvbsub)
If i want to change the container to mp4 i must transform dvbsub to
dvdsub.
For example:
ffmpeg -y -fix_sub_duration -i '/home/cavelier2/Videos/Irma la douce
 (VM).ts' -map 0:s -c:s dvdsub -f matroska sub.mkv or mp4.
I get the message: canvas_size(0:0) is too small(659:543) for render.
With french tnt i got some differents canvas according to differents
channels with the same code line:
canvas = 659:543 || canvas = 381:549 || canvas = 711:549 .
Question: how ffmpeg computes this canvas ?
Thanks.
-- 
cordialement jean-jacques. 
Ce message a été écrit sur un système libre mageia
"Tant que l'homme sera mortel, il ne pourra pas être totalement
décontracté..." Woody Allen.