FfmpegInputs.md

layout	title	nav_order	parent
default	Ffmpeg Frame Sequences Specification	2	Encoding Overview

Ffmpeg Frame Sequences Specification

Care needs to be taken when specifying frame sequences, since there is additional metadata that would normally be present in a movie container (e.g. mp4 or mov files) that are not present in an image sequence.

Ffmpeg will convert an image sequence using the image2 demuxer. This provides the code for determining how to wildcard as well as how to specify the frame-rate.

Image sequences

There are two approaches for defining image sequences, globbing and printf style.

Printf image sequence definition.

The conventional approach is to define the image sequence with a "%d" or a "%0Nd", which specifies where the frame number should go, for example: img.%d.png

would match: img.0.png img.1.png img.2.png ... img.10.png etc.

img.%04d.png means the numbers need to be zero padded to 4 digits, so it would match img.0000.png img.0001.png img.0002.png ... etc.

By default, the frame number is expect to start from 0, but you can define it with the flag: -start_number, e.g.:

ffmpeg -start_number 1 -i img.%04d.png foo.mov

Would start from frame number img.0001.png

If not defined the end frame will be the last continuous frame in the frame sequence, so if you have a missing frame it will stop there. You can define the number of frames to capture using the -frames:v flag.

N.B. In a windows command shell, % has a special meaning, so you may need to escape the "%", by replacing it with %%, or quote it, e.g.:

ffmpeg -start_number 1 -i img.%%04d.png foo.mov

TODO TEST.

Globbing image sequence definition.

There is a globbing option that makes it a little easier to specify a block of frames, since you dont need to specify the first frame.

ffmpeg -pattern_type glob -i "img.*.png" foo.mov

Will grab all frames that start with img. and end with ".png"

Frame Rate

The following aliases are defined for framerate values, either -r or -framerate parameters.

| ntsc | 30000/1001 | 29.97 fps equivalent. | | pal | 25| | | qntsc | 30000/1001 | VCD compliant NTSC | | qpal | 25 | VCD compliant PAL | | sntsc | 30000/1001 | square pixel NTSC | | spal | 25 | square pixel PAL | | film | 24 | | | ntsc-film | 24000/1001 | correct 23.98 |

Is is prefereable to use a fractional rate, or one of the above settings where possible, for example -r 30000/1001 is the more precise version of 29.97.

Other common fractional rates not defined with presets include:

| 60000/1001 | 59.94 fps equiavalent. | | 120000/1001 | 119.88 fps equivalent |

If not specified, the default framerate chosen is 25 fps (i.e. pal).

TODO Confirm no difference between -r and -framerate. -framerate seems a little more proper.

Looping

You can loop the input file with the -loop 1 parameter, e.g.:

ffmpeg -y -pattern_type glob -loop 1 -framerate ntsc -i "../sourceimages/chip-chart-1080-noicc.*.png" -pix_fmt yuv444p10le  -frames:v 100  ./chip-chart-yuvconvert/looptest.mov

Note, you want to control the number of frames to output, for a long sequence you would put the -frames:v 100 before the "-i" flag, but here we are putting it before the output, since we want it to apply to the overall looping input, not the input sequence.