This is #DS106

One of the best things about DS106 is the riffing and playing with other participants. I dipped my toe in the daily create stream yesterday and pulled out a great example:

As I saw this late there were already a great bunch of responses, I giffed:

tdc1713

I had intended to have the notes ping right off the sheet, but this looked ok and it was late (so late I lost a few frames that would have made the jump a little more dynamic).

Todd said:

and Ron:

Today Viv iced the cake:

There are deeper thinks going on on DS106 but this is #4life

This ain’t no Rotoscope

As usual it started with a tweet:

The referenced tweet looked interesting:

I’ve tried Rotoscoping before, but this looked good. I hied over to MacOS/OSX – Paint of Persia community – itch.io, read about it. paid $5 and set things up.

At that point I realise that the app helps you to manually trace frames, seemed a wee bit time consuming. I though I’d leave it for another day. I did start musing on doing something similar, I was thinging FFmpeg & ImageMagick.

Google takes me to Fred’s ImageMagick Scripts: CARTOON.

so the plan is:

  1. export a movie to a series of images.
  2. cartoonise those image
  3. create a move from the cartoon images

Bounus points for getting the sound track back in.

I already have ffmpeg & imagemagick installed on my mac. These are commandline tools.

I downloaded the cartoon script

First find a video:

YouTube downloader tool – Fastesttube!

Rename the downloaded file shower.mp4

Open the terminal and cd into the folder that has the cartoon script and the movie in it.

First extract lots of still images (make a folder shower first mkdir shower):

ffmpeg -i shower.mp4 -r 6 shower/out_%04d.jpg

this takes the input file (-i) at 6 frames per second and create jpg files in the shower folder with the file name out_0001.jpg ,out_0002.jpg ect

Given the movie is 3 minutes 22 seconds long I end up with 1212 images in the folder. I delete the last few manually to give me 1023 images.

I now need to loop through all of those images and create a cartoon version.

the use of cartoon is basically:

./cartoon face.jpg temp.jpg

There are some paramaters you can use but I stuck to the default.

This would take the image on the left and create the one on the right:

face.out

So I move into the shower folder cd shower

mkdir out;for i in *.jpg; do ../cartoon $i out/$i.jpg;done

this script:
makes a nother folder called out

for i in *.jpg for each file with a .jpg extension in the current folder do does:

../cartoon $i out/$i.jpg runs the cartoon script from the folder above (../), and saves the output file in the out folder with the same name.

When I kicked this off I quickly realised it would take a while to r=un through 1212 images, so went to bed.

The next morning I have 1023 cartoons (I don’t imagine that it took all night).
out_0672

Time to stich these together, it took me a few goes to get the paramaters right. (google helps).

ffmpeg -f image2 -framerate 6 -i out_%04d.jpg -c:v libx264 -pix_fmt yuv420p out.mp4

  1. -f chooses the format image2 which makes an image sequence.
  2. -framerate 6 to get the smae length of movie as I started with
  3. -i infiles confusinging names out_0001.jpg ect the %04d means look for 4 figured numbers starting with 0001
  4. -c codec
  5. -pix_fmt is the pixel format, I dn’t know much about these but the script failed until I added it in

It took me a few shots until I got this right.

I then moved the out.mp4 video to the same folder as the original and

ffmpeg -i shower.mp4 -acodec copy -vn shower-audio.mp4

To extract the audio from the original file.

then:

ffmpeg -i out.mp4 -i shower-audio.mp4 -vcodec copy -acodec copy -shortest final.mp4

To add the audio from the original to the out video to give me a final one.

the -shortest parameter gets rid of the audio at the end to make up for the frames I removed.

Bingo:

At this point I remembered iMovie has a comic filter…

This post is mostly as an aid to my occasional dip into the world of commandline video editing. Posting helps me remember. It also plays a wee bit fast and loose with copyright.

The featured image is a gif giffed from a few of the stills.

ps this is quite a disturbing clip, I didn’t really watch it till I finished, could have picked a nice one!

Update, Ron commented:

What would it look like if you’d run the result videos through the script one more time. Keeping the same number of jpegs but making the lines more stand out and the fills less so you’d have black outlines and white?

That didn’t make much difference os I did some tests with the parameters for cartoon, and Ron went with increasing the brightness: ../cartoon -b 300

This gave me this:

Accidental Allure

In the past I’ve made a few experiments with randomly layering and combining images: Glen Finlas -evaluate-sequence subtract and Averages (The Prisoner) for example.

A couple of weeks ago I started playing with combining images in the browser. There are several ways to go about this, I found a nice script to blend two images on a canvas and gave that a shot. It worked well and gave interesting results.

I though that using the Flickr API I could gather a list of images and randomly blend them two at a time.

Flickr’s API will return a json list. I started using the flickr.interestingness.getList which produced some interesting (sic) combinations. However when I started to get the license of the photos most were not labled for reuse.

I switched to using a standard search (flickr.photos.search ) which allowed me to search for license that allowed reuse.

I also switched to using CSS and background-blend-mode, this allows you to have multiple images set on a background and blend them.

For example using these images:

And this code:

<div style="border:solid 1px;width:500px;height:400px;background-image:url('https://c2.staticflickr.com/8/7587/26482589423_daa3bbdbd1.jpg'),url('https://c2.staticflickr.com/8/7012/6677861899_ef6e012bc8.jpg');background-blend-mode:multiply;"></div>

give me this:

With this in my toolbag I could pull in a flickr feed, extract image URLs and info about each photo and randomly combine them. They are displayed for 10 seconds each.

This gave me this:

Random Flickr Blendr

Here are a few random blends, screen captured:

I’ve found the pictures quite compelling.

On interesting this was the change I noticed when I swapped from the interestingness list to a search for creative commons images ordered by interestingness-desc. The images became more subtle and less HDRish, i think thy are more interesting and less glossy. An unusual win for Creative commons.

Over in DS106 land the page was used for a daily create:

#tdc1588 Turn @johnjohnston’s Random Image Pairing into a Self-Help Book Title | The (new) Daily Create

Which turned up some nice images and a fairly crazy bunch of titles.

My Own:

tdc1588

What was also interesting was some responses to the page:

So I an quite pleased with the result of this bit of experimenting. I’ve learnt a little more about CSS, images, JavaScript and even practised a bit of git. On the git front I’ve installed ezyang/git-ftp which is a quick and efficient way of pushing changed files to a website via FTP and works very well indeed. Saves working directly on line or opening an FTP application.

GifDub

gif-dub-screenshot

GifDub

Using Neocities who say We provide free web hosting and tools that allow anyone to make a website. and Neocities will never sell your personal data or embed advertising on your site..

GifDub is a sort of sound and gif mashup using the Giphy Api and Freesound API. The idea is to play several random gifs and audio files at the same time. You can toggle sounds on and off, replace gifs and sounds.

It is fairly silly but has been a lot of fun to put together. I suspect it will not work in Internet Explorer.

A work in progress, I’d like to add some more features and get it working in Internet Explorer.

This is not part of any ds106 assignment but it feels like DS106 to me, and it has gifs #wejamecono.

A #DS106 DailyCreate Twitter Bot

dailycreate-botThe DailyCreate Bot

I’ve always liked random and automated random things. While these are not strictly part of storytelling I’ve managed to bring them into DS106 whenever I can.

A while back I set up @DailyCreateBot for some reason or other. Obviously a Twitter bot of some kind to do with the Daily Create. I do remember having trouble with the OAuth requirements of the more recent Twitter API and giving up.

Last weekend, on a rainy day I blew the dust of my raspberry pi and got it online and set up as a server. I was not too sure what to do with it at the time.

During the week I did revisit a project to use the pi to flash some lights depending on a Twitter search. I don’t have hardware for that but I was interested in how simple the project was. There seems to be plenty of libraries that can sort out Authentication to Twitter for you now. A bit of googling and thinking, mostly googling and I have a Twitter bot set up.

The @DailyCreateBot will reply with a suggestion of a photo challenge of you mention him on Twitter. I am using the same list that Alan Levin provided for me for the photoblitz.

The @DailyCreateBot runs on Python. This is where the pi comes in I would not even know where to begin to find out how to host a python app but the pi lets me do that easily.

I am not proposing to write a step-by-step guide here but it is worth mentioning that several things went wrong or did not work as expected. All were beyond my 2 weeks worth of Python on the mechanical mooc . All were solved by a wee bit of googling and sometime just repeating things till they worked. The delight of working on a pi is that I knew I’d I totally messed up I could just reformat the SD card, install an so again and be back to square two.

I had already:

  • installed one of the basic OS FOR THE PI
  • Set up SSH access so that I can get ‘on’ to the pi from the terminal application on a mac and via SSH apps on iOS.
  • set up the pi as a web server and sorted out the DNS

Next:

*I found a python library and example code that replied.
*I added logic to reply with a random string taken from a list of challenges.
*Tested it a bit.

Then I posted to the DS106 Google + group and a few kind folk tested it a bit. Rochelle asked:

That is cool +John Johnston . It worked for me right out of the bot box. Do we upload to Twitter, tag them DailyCreateBot? I’d like to see what others have done. 🙂

Which got me thinking. A quick google found a php/JavaScript solution to showing tweets with the hashtag #dailycreatebot and I’ve got this up and running.

All very much a work in progress. There are few things to be ironed out:

  • the Python bot falls over every now and again complaining about UTF8 I need to google that some more.
  • the web page showing images just uses the styles used in the demo of the code. I need to tidy it up and perhaps skip tweets with the hashtag but no images.
  • there is also the problem Rochelle pointed out that if you reply to the bot you get another prompt. I wonder if I could turn off replies if there is an image in the tweet?

Anyway if your expectations are low you can join in:

  1. Tweet @DailyCreateBot and get a prompt.
  2. Tweet your photo with the hashtag #DailyCreateBot
  3. See what other folk are doing.
  4. Let me know of any interesting problems.