Problem padding with ffmpeg

[joesphroth]

Hi I need to convert some video to tiff images for conversion later for digital cinema. The tiffs need to be 1920x1080 and up untill now all the video I have converted has been in that resolution.
The video I have now is h264 and is 1920x1040, so I need to add 20 pixels of black to the top and bottom. I have tried the padding and scale filter and I always get unsupported output error. I'm not sure if my syntax is wrong. (I'll post it later I'm at work right now)
I don't want to reencode the video just output to raw tiff. Is there a better way to do this?
Any help would be great.
Thanks

[poisondeathray]

not sure about the ffmpeg syntax for adding borders , but it's pretty easy to do in avisynth

e.g

FFVideoSource("video.ext")
AddBorders(0,20,0,20)
ConvertToRGB24()
ImageWriter("image%04.tiff")

Open that .avs in vdub and use file=> run video anaysis pass and it will write the tiff's , image0000, image0001, etc... You can adjust the name syntax or placeholder digits

[joesphroth]

Thanks a lot I'll try this as soon as I can.

[joesphroth]

I couldn't get the avs to work in virtualdub but i was able to run the avs (with just the video source and borders) through ffmpeg to get the tiffs.

Thanks a lot.

[poisondeathray]

Cheers

Just a note about YUV=>RGB conversions in ffmpeg - I think swscale will use Rec601 matrix, so that might not be correct for HD footage when converting to TIFF (colors will be slightly shifted)

In avisynth you can specify Rec709 which is usual matrix for HD footage

FFVideoSource("video.ext")
AddBorders(0,20,0,20)
ConvertToRGB24(matrix="rec709")

You can use Trim(x,y) to do test out a few frames and see the difference

e.g. Trim(0,100) would return frames 0-100 in avisynth

[joesphroth]

EDIT: double post

[poisondeathray]

Yes, I understand what you're saying, I'm not talking about vdub.

If you don't add that 3rd line, ffmpeg swscale will do the YUV=>RGB conversion with Rec601 , so your TIFF's colors might be slightly off. By adding that 3rd line, you take control of the conversion instead of swscale, and uncompressed RGB is frameserved into ffmpeg

Test out a short section with and without, there are differences. Normal HD workflow uses ITU Rec.709 for everything
http://en.wikipedia.org/wiki/Rec._709
http://en.wikipedia.org/wiki/Rec._601

[joesphroth]

OK thanks for the input I'll use rec.709. The TIFFs are converted to XYZ JPEG2000 files and compiled to a Digital Cinema Package. The TIFFs are 24bit and very big (about 6mb each) is there anyway to make them smaller without quality loss?

[poisondeathray]

You can use lossless compression like PNG instead of TIFF, but what are you using for the DCP conversion ? Open Cinema Tools ?

FFMpeg will convert TIFF to 8bpc or 24bit total (8 red, 8 blue, 8 green)

Open source / free workflows usually convert that 8bit PNG or TIFF input to 12bit TIFF (that's 12 bpc, bits per channel, or 36 bit total) with imagemagick - even larger filesizes - before the j2k conversion

See this:
http://bitfilms.blogspot.com/2010/11/making-dcp-entirely-with-open-source.html