[FFmpeg-user] How to use ocr filter

Moritz Barsnick barsnick at gmx.net
Thu Sep 17 17:28:00 CEST 2015

On Thu, Sep 17, 2015 at 15:12:17 +0000, Paul B Mahol wrote:
> On 9/17/15, nicolab <robelt2525 at gmail.com> wrote:
> > When I using ocr filter, how to output ocr text file ?

> drawgraph accepts only floats values.
> ffplay ~/img.png -vf "ocr,split[ocr][o1],[ocr]lutyuv=y=0:u=128:v=128,drawtext=fontcolor=white:x=10:y=10:text=%{metadata\\:lavfi.ocr.text}[o2],[o1][o2]vstack"

Whatever you're doing, Paul, isn't outputting to text file, as nicolab
requested. ;-)

I think ffprobe is the correct approach. It can output - as text - the
metadata which the ocr filter inserts.

I can't test, since none of my binaries has ocr support, but probably
something like:

$ ffprobe -show_frames -f lavfi -i "movie=img.png,ocr"

and some additional options for selecting the correct fields.

ffmpeg also has an output format "ffmetadata", that might help as well
(but I guess it contains _all_ metadata, and fields can't be selected
or filtered).


More information about the ffmpeg-user mailing list