I've recorded a .gif of the screen with ffmpeg. I've used gifsicle and imagemagick to compress it a bit, but it's still to big. My intent is making it small by removing, say, a frame every 2 frames, so that the total count of frames will be halved.
I couldn't find a way to do it, neither with gifsicle nor with imagemagick. man pages didn't help.
How can I remove a frame from a .gif animation every n frames?
There is probably a better way to do it, but here is what I would do
First, split your animation in frames
convert animation.gif +adjoin temp_%02d.gif Then, select one over n frames with a small for-loop in which you loop over all the frames, you check if it is divisible by 2 and if so you copy it in a new temporary file.
j=0; for i in $(ls temp_*gif); do if [ $(( $j%2 )) -eq 0 ]; then cp $i sel_`printf %02d $j`.gif; fi; j=$(echo "$j+1" | bc); done If you prefer to keep all the non-divisible numbers (and so if you want to delete rather than to keep every nth frame), replace -eq by -ne.
And once you done it, create your new animation from the selected frames
convert -delay 20 $( ls sel_*) new_animation.gif You can make a small script convert.sh easily, which would be something like that
#!/bin/bash animtoconvert=$1 nframe=$2 fps=$3 # Split in frames convert $animtoconvert +adjoin temp_%02d.gif # select the frames for the new animation j=0 for i in $(ls temp_*gif); do if [ $(( $j%${nframe} )) -eq 0 ]; then cp $i sel_`printf %02d $j`.gif; fi; j=$(echo "$j+1" | bc); done # Create the new animation & clean up everything convert -delay $fps $( ls sel_*) new_animation.gif rm temp_* sel_* And then just call, for example
$ convert.sh youranimation.gif 2 20 
.gifs cannot be read by imagemagick as they are corrupted! They work fine in a browser, but imagemagick complains. I've tried recording directly to a .gif and recording to a .mkv and converting afterwards. Still corrupted - is there any way I can fix this "corruption" the gif has? identify; it will (hopefully) tell you which frames are corrupted. Then, for example, you could do something like that if it is the 42th frame that is corrupted: convert 'animation.gif[0-41]' anim_nocorrupt_1.gif and convert 'animation.gif[42-99]' anim_nocorrupt_2.gif and then merge the produced files. There is an alternate version of MBR's answer, as a bash function:
gif_framecount_reducer () { # args: $gif_path $frames_reduction_factor local orig_gif="${1?'Missing GIF filename parameter'}" local reduction_factor=${2?'Missing reduction factor parameter'} # Extracting the delays between each frames local orig_delay=$(gifsicle -I "$orig_gif" | sed -ne 's/.*delay \([0-9.]\+\)s/\1/p' | uniq) # Ensuring this delay is constant [ $(echo "$orig_delay" | wc -l) -ne 1 ] \ && echo "Input GIF doesn't have a fixed framerate" >&2 \ && return 1 # Computing the current and new FPS local new_fps=$(echo "(1/$orig_delay)/$reduction_factor" | bc) # Exploding the animation into individual images in /var/tmp local tmp_frames_prefix="/var/tmp/${orig_gif%.*}_" convert "$orig_gif" -coalesce +adjoin "$tmp_frames_prefix%05d.gif" local frames_count=$(ls "$tmp_frames_prefix"*.gif | wc -l) # Creating a symlink for one frame every $reduction_factor local sel_frames_prefix="/var/tmp/sel_${orig_gif%.*}_" for i in $(seq 0 $reduction_factor $((frames_count-1))); do local suffix=$(printf "%05d.gif" $i) ln -s "$tmp_frames_prefix$suffix" "$sel_frames_prefix$suffix" done # Assembling the new animated GIF from the selected frames convert -delay $new_fps "$sel_frames_prefix"*.gif "${orig_gif%.*}_reduced_x${reduction_factor}.gif" # Cleaning up rm "$tmp_frames_prefix"*.gif "$sel_frames_prefix"*.gif } Usage:
gif_framecount_reducer file.gif 2 # reduce its frames count by 2 
local orig_delay=$(gifsicle -I "$orig_gif" | sed -ne 's/.*delay \([0-9.]\+\)s/\1/p' | sort | uniq -d | head -n 1) The best way to do it would be with ffmpeg, preferably when you initially create the gif rather than after for the sake of preserving quality.
There is almost no loss of quality thanks to the palettegen/paletteuse filter used (and the trick of setting it to not use a dither pattern), however it can lead to some color information getting lost, but it's better than the alternative which will always apply heavy dither patterns through sws_dither which is extremely damaging for the quality.
As an example here is a 24fps gif exactly 1 second long (so 24 frames). 
ffmpeg -i in.gif -vf "fps=12,split[s0][s1];[s0]palettegen=max_colors=256:reserve_transparent=0:stats_mode=diff[p];[s1][p]paletteuse=dither=none" out.gif 
And now it is still exactly 1 second long, but only 12 frames.
ffmpeg -r 48 -i in.gif -vf "fps=24,split[s0][s1];[s0]palettegen=max_colors=256:reserve_transparent=0:stats_mode=diff[p];[s1][p]paletteuse=dither=none" out.gif 
And this time it's only half a second long, with 12 frames.
For better context of the question.
fps=24/2fps=24/1.5fps=24/1.33333333333fps=24/1.25What you do is divide the source framerate by the n for every nth frame you wish to remove. So 24/3=8.
Then you subtract that value from the source framerate. 24-8=16
So fps=16 would remove every 3rd frame. There's probably a better way to calculate this but i'm not very much of a math guy so this is what I've got.
