###More fun with images

**WARNING**, If you are on a slow connection, go enjoy a cup of
coffee because it will take a few minutes for the animated images to
load.

I thought I might take the image reduction experiment one step
further and see what I could do with video.  Video over a
slow connection is an exercise in patience, but after giving some
thought, it might be possible albeit on a very limited scale.

![nuclear explosion](http://melton.sdf-us.org/images/explosion.gif)

I chose to take a similar approach that I took with still images
utilizing an ordered dithering algorithm.  Right now, I have had to
settle for an animated gif, but I suspect there is most likely
another more efficient format.  It will do the job for now as I
research further. In this case, I took a 10 second snippet [from
this video](https://youtu.be/m6DsC6Mmh34) which is a frighteningly
familiar scene.  The original video extract was 17 megabytes...waaay
too much for a slow connection (such as dialup) and the energy
footprint not only on the server and browser sides, but also
transport made the original snippet a budget buster. So, let us do
some more processing with the segment.  First I attempted to process
it with Imagemagick like I did with still images as Imagemagick can
output animated gifs.  The Atom processor on this old Toshiba choked
and then barfed, so no go with that method.  The next method
involved splitting out the individual frames with ffmpeg and then
converting each individual frame with Imagemagick.  This yielded a
total of 100 frames (10 seconds at 10 frames per second)  The
processor was much happier with this approach.  The first pass
without any ordered dithering algorithm produced a much smaller
animated gif at 2.7 megabytes.  Much better, but still not good
enough.  The next run, I applied some ordered dithering which
brought the file size to 324 kilobytes which is now in the realm of
where we need it.  I gave the snippet one last pass tweaking the
ordered dithering a bit and finally brought the final file size down
to 304k.  It still takes a couple of minutes to download over
dialup, but it's a start.  For those interested, here's the code:

`ffmpeg -i input -vf scale=320:-1:flags=lanczos,fps=10 \
	frames/ffout%03d.png`

\# and then process the frames

`convert -quiet -loop 0 -delay 10  \ 
	frames/ffout*.png -ordered-dither o4x4 +map \
	output.gif`

There is probably scope to tweak the scale a bit smaller, but one
can get the idea.  Of course there is a newer compression algorithm
called webp that does an admirable job and might be worth exploring,
but unfortunately the resulting file at 585k, it is still a bit too
large for a slow connection and would also consume more energy
to compose, serve, transport and display (particularly on legacy
browsers).  

![webp explosion](http://melton.sdf-us.org/images/explosion-2.webp)

Even the animated gif at the top, took two and half minutes to
download over a simulated dialup connection. Anyway, it's an
interesting exercise, but most likely I will confine this site to
still images to keep things lean and mean.  As I had mentioned in a
previous post, the Gopher protocol still has by far the tiniest
energy footprint as it is just plain ascii.

Gopher, it seems maybe garnering renewed interest not only in the
niche retro computing community, but perhaps also with some who are
weary of the web cruft and intrusiveness and who are looking for
just pure content.  Bleeding edge hardware all the way back to your
atticware [68k Mac can browse gopher since it makes very little
demand on the client machine.](https://tinyurl.com/ych8fqa9)  A few
web sites have responded to this resurgence such as [Project
Gutenberg](gopher://dante.pglaf.org/),
[Gopherpedia](gopher://gopherpedia.com/),
[Metafilter](gopher://gopher.metafilter.com/) and I'm sure there are
others.  Even some news portals are at least providing some text
only interfaces to their full sites.  [NPR](https://text.npr.org/)
comes to mind even though I'm not a fan.