If the purpose of covers is to visually differentiate among the set of books which are visible in a skeumorphic "bookshelf", could pre-computed image similarity metrics be used to maximize the difference between neighboring titles?
I specially liked the illustration ones. I wonder if perhaps some database of public art illustrations could be somehow consulted to generate covers for the books without illustrations?
I applaud this effort from the sense that generated book covers all seem to be so ugly and undesigned. There's nothing wrong with a generic, but well designed book cover. I mean, how many virtually identical hardbound books are there sitting on people's bookshelves...and they look beautiful instead of ugly.
Some kind of more clever generative art I think is called for. It should even have some kind of sensible themeing, like all books from the same author look kind of alike or something.
Yeah, interesting! It'd be cool if you had a generative book cover that was fed the author name (and maybe some other information) as a seed, where the function gives output from the same author to have the same colour, or patterns, or something, that created images similar to the gold leaf patterns. That's actually sounding quite interesting, I might have a go at it!
Both cover design and interior layout are insanely challenging. I run a small press (everybody needs a hobby: http://www.siduri.net) and the amount of time spent on covers is way out of proportion to everything else.
In an ideal world, the artist has to a) understand the book and b) create a cover that sums it up and c) produce an alternative version that isn't too heavy for the e-book.
The difference between e-book requirements and print are huge, and a few of the recent Penguin e-books I've bought all have the same cover (the Penguin logo) which looks to me like the editors just punted on the issue.
So I applaud this effort. The problem they are solving is much more difficult than it looks from the outside, and they are doing an impressive job even though the results may look less than ideal to some. Things like Project Gutenberg would be considerably enhanced if they had decent covers, and any scheme for auto-generating them that isn't positively awful [] would really help.
[] I've gotta say the first purely geometric covers in the article aren't quite my cup of tea.
Made me remember a series of books from '70s-'90s in Poland (about computer programming), somewhat similar in style, although I thought they were more so before googling. In case you'd be interested:
On first glance, those look almost like an Arabic letter. I suppose that for more abstract topics like programming, an abstract-looking cover is pretty common.
"For their research, the team chose 1,700 paintings by 66 artists, covering the 15th to the late 20th century. Using a technique that analyzes visual concepts called "classemes"—wherein objects, color shades, subjects' movement, and more are marked—the researchers created a list of 3,000 classemes for each painting, data which The Physics arXiv Blog compares to a vector. Then, they used an artificially intelligent algorithm to evaluate the vectors and look for similarities or overlapping qualities among the 1,700 paintings. ArXiv adds, "To create a ground truth against which to measure their results, they also collate expert opinions on which these artists have influenced the others."
I wonder if it would be possible to generate a cover based on the plot?
Character name detection is pretty easy
Gender detection for the characters depends on the year of the book, but is possible with reasonable accuracy (although one would need to be careful of translations).
Key scene detection in text isn't something I know much about. I've read some research on recognising emotion in text, and it might be possible to use something like that on a rolling basis to find the most important parts of the book.
Using that information to create an illustration... would be challenging.
I can imagine algorithms to draw the character by extracting descriptions from the book. Putting them into the context described by some key scene would have a few additional problems.
29 comments
[ 6.0 ms ] story [ 80.7 ms ] threadFor ebooks with images [Gutenberg]: https://github.com/mgiraldo/tenprintcover-p5
This also reminds me of the autogenerated covers on the (ridiculously overpriced) books containing only Wikipedia articles, which were equally generic-looking but still somewhat relevant (e.g. https://images.our-assets.com/fullcover/2000x/9786131724619.... , http://www.amazon.com/Aisle-Frederic-P-Miller/dp/B009CKDZII/ )
Of course, I think the saying "Don't judge an eBook by its cover" also applies...
102 degrees would have been more sensible, while providing sufficient contrast.
Using a different, more perceptually-based colour space would be even better.
https://github.com/mgiraldo/tenprintcover-p5
This other project has some interesting geometric explorations:
http://variable.io/generative-covers/
I hereby coin the term "depressionism" - you heard it here first, folks!
might be a good start. These are images auto-extracted from public domain books at the Internet Archive.
http://bibliotique.net/images/080511-aa-1.jpg
http://www.dargate.com/cat/248_auction/248_images/1722.jpg
The difference is that there is some sense of trying to make them at least a little eye pleasing. Decent fonts, some pretty gold leaf designs.
This is not good enough http://i.imgur.com/YqjnSsF.png
Some kind of more clever generative art I think is called for. It should even have some kind of sensible themeing, like all books from the same author look kind of alike or something.
http://graywolfseo.com/wp-content/uploads/library-shelf.jpg
I'm not sure I like how the examples provided here look especially, but I think the idea is a good one.
http://sprachkreis-deutsch.ch/2013/06/30/reclam-verlag-folgt...
http://www.phantasienreisen.de/2012/02/19/reclam-klassiker-i...
Book wall: http://commons.wikimedia.org/wiki/File:Reclam_Universalbibli...
red: classics in their original foreign language
orange: bi-lingual books
blue: study guides
green: original historical sources with explanations
magenta: non-fiction (politics, history, society, natural sciences, art, music or religion)
source: https://en.wikipedia.org/wiki/Reclam
@paulg
When I got rich I let myself buy nice editions of books. But I notice the books I read are mostly cheap paperbacks.
In an ideal world, the artist has to a) understand the book and b) create a cover that sums it up and c) produce an alternative version that isn't too heavy for the e-book.
The difference between e-book requirements and print are huge, and a few of the recent Penguin e-books I've bought all have the same cover (the Penguin logo) which looks to me like the editors just punted on the issue.
So I applaud this effort. The problem they are solving is much more difficult than it looks from the outside, and they are doing an impressive job even though the results may look less than ideal to some. Things like Project Gutenberg would be considerably enhanced if they had decent covers, and any scheme for auto-generating them that isn't positively awful [] would really help.
[] I've gotta say the first purely geometric covers in the article aren't quite my cup of tea.
Worth noting our effort is mostly a side project that supports the _real_ project which is the ebook reader app
If no cover is available, I would much prefer a simple text label.
When I do find such a book, I waste no time in stripping off the cover.
Putting a fake cover on an old book is akin to vandalism.
http://imgur.com/JvhuomK
http://imgur.com/Ktnec4l
Also similar, this set of famous theoretical physics books: http://2.bp.blogspot.com/-sj0HK_o3hD0/TlhDu7sl_FI/AAAAAAAAAh...
Is there software which models the human perceptual system, e.g. algorithms that have been trained on fanous art?
Related: http://thecreatorsproject.vice.com/blog/an-intelligent-algor...
"For their research, the team chose 1,700 paintings by 66 artists, covering the 15th to the late 20th century. Using a technique that analyzes visual concepts called "classemes"—wherein objects, color shades, subjects' movement, and more are marked—the researchers created a list of 3,000 classemes for each painting, data which The Physics arXiv Blog compares to a vector. Then, they used an artificially intelligent algorithm to evaluate the vectors and look for similarities or overlapping qualities among the 1,700 paintings. ArXiv adds, "To create a ground truth against which to measure their results, they also collate expert opinions on which these artists have influenced the others."
Character name detection is pretty easy
Gender detection for the characters depends on the year of the book, but is possible with reasonable accuracy (although one would need to be careful of translations).
Key scene detection in text isn't something I know much about. I've read some research on recognising emotion in text, and it might be possible to use something like that on a rolling basis to find the most important parts of the book.
Using that information to create an illustration... would be challenging.
I can imagine algorithms to draw the character by extracting descriptions from the book. Putting them into the context described by some key scene would have a few additional problems.
http://www.tapor.ca/?id=349
http://www.businessinsider.com/apple-buys-booklamp-2014-7